Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Temperament, parenting, and the development of childhood obesity Hejazi, Samar 2007

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2008_spring_hejazi_samar.pdf [ 775.73kB ]
Metadata
JSON: 24-1.0066157.json
JSON-LD: 24-1.0066157-ld.json
RDF/XML (Pretty): 24-1.0066157-rdf.xml
RDF/JSON: 24-1.0066157-rdf.json
Turtle: 24-1.0066157-turtle.txt
N-Triples: 24-1.0066157-rdf-ntriples.txt
Original Record: 24-1.0066157-source.json
Full Text
24-1.0066157-fulltext.txt
Citation
24-1.0066157.ris

Full Text

  TEMPERAMENT, PARENTING, AND THE DEVELOPMENT OF CHILDHOOD OBESITY    by SAMAR HEJAZI M.N., Emory University, 1994 B.S.N., Jordan University of Science & Technology, 1991   A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIRMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSPHY   in  THE FACULTY OF GRADUATE STUDIES (Nursing)  THE UNIVERSITY OF BRITISH COLUMBIA November 2007  © Samar Hejazi, 2007  ii Abstract The purpose of this study was two-fold: (a) to identify, in a large representative sample of Canadian children, the age-related trajectories of overweight and obesity from toddlerhood into childhood and (b) to investigate the associations between these trajectories and children’s temperaments, their parents’ parenting practices and their interactions. Potentially important familial characteristics (i.e., the parents’ or surrogates’ age, income level, and educational attainment) were considered in the models. The sample for this study was drawn from the Canadian National Longitudinal Survey of Children and Youth (NLSCY). Group-based mixture modeling analyses were conducted to identify the number and types of distinct trajectories in the development of obesity (i.e., to explicate the developmental processes in the variability of childhood obesity) in a representative sample of children who were between 24 to 35 months of age, at baseline, and followed biennially over a 6-year span. Discriminant analysis was conducted to assess the theoretical notion of goodness-of-fit between parenting practices and children’s temperament, and their association with membership in the BMI trajectory groups. The results of the group-based modeling established three different BMI trajectories for the boys, namely: stable-normal BMI, transient-high BMI, and j-curve obesity. The analyses revealed four different trajectories of BMI change for the girls: stable-normal BMI, early-declining BMI, late-declining BMI, and accelerating rise to obesity. The multivariate analysis revealed that the combined predictors of the obesity trajectories of the girls (group membership) included having a fussy temperament, ineffective parenting, and parents’ educational attainment. Predictors of the boys’ obesity trajectory (group membership) included household income, parental education, and effective parenting  iii practices. Understanding the different ways in which a child may develop obesity will allow nurses and other health professionals to take different approaches in the assessment, intervention and evaluation of obesity and obesity-related health problems. The results of this study further our understanding of factors associated with the development of obesity at a young age and hence may inform the development of early preventive programs.                   iv Table of Contents  Abstract ........................................................................................................................................... ii Table of Contents........................................................................................................................... iv List of Tables .................................................................................................................................. x List of Figures .............................................................................................................................. xiii Acknowledgements...................................................................................................................... xiv Chapter 1: Introduction ................................................................................................................... 1 Chapter 2: Literature review and research questions...................................................................... 7 Risk factors associated with childhood obesity .......................................................................... 7 Genetics and early life experiences......................................................................................... 8 Physical inactivity ................................................................................................................. 10 Prevention of obesity ................................................................................................................ 11 Temperament and obesity ......................................................................................................... 12 Parenting and obesity................................................................................................................ 17 Parenting and temperament....................................................................................................... 26 Potential pathways to obesity.................................................................................................... 32 Temperament and maternal feeding practices ...................................................................... 32 Potential confounders: Parents’ education, age and income..................................................... 34 Summary of the research problem ............................................................................................ 36 Research purpose ...................................................................................................................... 38 Research questions.................................................................................................................... 38  v Chapter 3: Methods....................................................................................................................... 40 Data source and sample ............................................................................................................ 40 Study cohort .......................................................................................................................... 44 Data collection .......................................................................................................................... 44 Measures ................................................................................................................................... 46 BMI: Obesity and overweight............................................................................................... 46 Conceptual definition: BMI .............................................................................................. 46 Operational definitions: BMI, overweight and obesity..................................................... 49 Temperament ........................................................................................................................ 50 Conceptual definition........................................................................................................ 50 The temperament scale ......................................................................................................... 53 Operational definition: Difficult/fussy temperament............................................................ 55 Measures of fussy/difficult temperament.............................................................................. 56 Scoring .................................................................................................................................. 59 Parenting practices .................................................................................................................... 59 Conceptual definition............................................................................................................ 59 Parenting scales..................................................................................................................... 60 Measures of parenting........................................................................................................... 61 Scoring .................................................................................................................................. 66 Study design.............................................................................................................................. 67 Data organization and analysis ............................................................................................. 67 Growth curve models ............................................................................................................ 68 The model: A finite mixture model ...................................................................................... 69  vi Model selection..................................................................................................................... 72 Optimal number of groups ................................................................................................ 72 Specifying the polynomial function.................................................................................. 75 Probability of group membership ..................................................................................... 75 Calculation of posterior group membership probabilities ................................................ 76 Model adequacy .................................................................................................................... 77 Diagnostic 1: The average posterior probability of assignment ....................................... 77 Diagnostic 2: The odds of correct classification............................................................... 78 Diagnostic 3: Confidence intervals for the predicted point estimates of the trajectories . 78 Diagnostic 4: Estimated group probabilities versus the proportion of the sample assigned to the group ........................................................................................................ 79 Discriminant analysis............................................................................................................ 79 Variable selection and model building ............................................................................. 80 Missing data .......................................................................................................................... 80 Chapter 4: Results ......................................................................................................................... 82 The participants (study cohort) ................................................................................................. 82 Attrition analysis....................................................................................................................... 84 Sample description.................................................................................................................... 85 Geographical characteristics of the participants ................................................................... 86 Demographic characteristics of the responding parents (surrogates) and their spouses....... 87 The children’s health status .................................................................................................. 90 Univariate analysis.................................................................................................................... 92 BMI ....................................................................................................................................... 92  vii Temperament ........................................................................................................................ 96 Parenting ............................................................................................................................... 98 BMI trajectory modeling results ............................................................................................... 99 Data organization and analysis ............................................................................................. 99 Model selection..................................................................................................................... 99 The optimal number of groups........................................................................................ 100 Summary of the optimal number of groups .................................................................... 104 Specifying the polynomial function................................................................................ 104 Proportion of individuals in each of the groups.................................................................. 108 The calculation of posterior group membership probabilities ............................................ 109 Model adequacy .................................................................................................................. 110 Diagnostic 1: The average posterior probability of assignment ..................................... 110 Diagnostic 2: The odds of correct classification............................................................. 112 Diagnostic 3: Calculating confidence intervals for the predicted point estimates of the trajectories of BMI.......................................................................................................... 113 Diagnostic 4: The estimated group probabilities versus the proportion of the sample assigned to each group.................................................................................................... 116 A description of the groups based on Cole’s classifications for BMI ................................ 118 A description of the groups’ BMI trajectories .................................................................... 121 Potential confounders: Parents’ (surrogates’) education, age and income ............................. 125 Bivariate relationships between trajectory group membership and the variables of primary interest..................................................................................................................................... 129 Discriminant function analysis ............................................................................................... 133  viii A brief summary of the major results ..................................................................................... 139 Chapter 5: Discussion ................................................................................................................. 142 General overview.................................................................................................................... 142 Summary of the findings......................................................................................................... 145 Prevalence of overweight and obesity ................................................................................ 145 Research question 1 ............................................................................................................ 146 Research question 2 ............................................................................................................ 147 Research question 3 ............................................................................................................ 147 Research question 4 ............................................................................................................ 148 Connections and parallels with the existing literature ............................................................ 149 Prevalence of overweight and obesity ................................................................................ 149 BMI trajectories .................................................................................................................. 151 Temperament and BMI trajectories: Bivariate associations ............................................... 154 Parenting and BMI trajectories: Bivariate associations ...................................................... 155 Theoretical goodness-of-fit ................................................................................................. 160 Confounding variables ........................................................................................................ 162 Confidence in the findings ...................................................................................................... 163 Gender differences .................................................................................................................. 163 Gender variability in temperament and obesity .................................................................. 163 Gender variability in parenting and obesity........................................................................ 165 Limitations and implications................................................................................................... 167 Methodological limitations ................................................................................................. 167 The validity of the parenting practices and temperament scales .................................... 168  ix The validity of parental reports of children’s weight and height.................................... 170 Additional potential biases and confounders .................................................................. 171 Implications......................................................................................................................... 172 Implications for practice ................................................................................................. 173 Implications for future research: Measurement implications for parenting practices and temperament............................................................................................................. 174 Methodological implications for future research............................................................ 175 Longitudinal research and additional contributing risk factors ...................................... 176 Conclusion .............................................................................................................................. 176 References................................................................................................................................... 178 Appendix A................................................................................................................................. 216  x List of Tables  Table 1. Summary of the samples included  in each survey cycle ............................................... 44 Table 2. Cole's obesity and overweight cut-offs........................................................................... 50 Table 3. Age-specific measures of fussy/difficult temperament (24-35 months)......................... 56 Table 4. Positive interaction and scale rating ............................................................................... 61 Table 5. Consistency in discipline and scale rating ...................................................................... 63 Table 6. Ineffective parenting and scale rating............................................................................. 64 Table 7. Age groups, survey cycles and variables measured........................................................ 67 Table 8. Interpretation of 2loge(B10) ............................................................................................. 74 Table 9. Cases with outlier BMI values per survey cycle............................................................. 83 Table 10. Counts of cases with acceptable BMI values................................................................ 84 Table 11. Geographical characteristics ......................................................................................... 86 Table 12. Parents’ (surrogates’) demographics ............................................................................ 88 Table 13. Children’s health status................................................................................................. 91 Table 14. Descriptive statistics for the girls’ and boys’ BMI....................................................... 93 Table 15. Prevalence rates of overweight and obesity based on Cole's cutoffs............................ 95 Table 16. Temperament: Descriptive statistics............................................................................. 97 Table 17. Parenting: Descriptive statistics.................................................................................... 98 Table 18. Selecting the number of groups based on the BIC (girls)........................................... 100 Table 19. Selecting the number of groups based on the BIC (boys) .......................................... 102 Table 20. Interpretation of the 2loge(B10) ................................................................................... 102 Table 21. Tabulated BIC and 2loge(B10) values for the girls' models......................................... 103 Table 22. Tabulated BIC and 2loge(B10) for the boys' models ................................................... 103  xi Table 23. BIC calculations for different models with combinations of alternative specifications for the girls' data........................................................................................... 106 Table 24. BIC calculations for different models with combinations of alternative specifications for the boys' data .......................................................................................... 108 Table 25. Probability of group membership ............................................................................... 109 Table 26. Average assignment probability conditional on assignment by maximum posterior probability rule for the girls’ model.................................................................................... 111 Table 27. Average assignment probability conditional on assignment by maximum posterior probability rule for the boys’ model ................................................................................... 111 Table 28. Odds of correct classification for the girls’ model ..................................................... 113 Table 29. Odds of correct classification for the boys’ model ..................................................... 113 Table 30. Mean BMIs and 95% CIs by the girls’ trajectory groups ........................................... 114 Table 31. Mean BMIs and 95% CIs by the boys’ trajectory groups........................................... 115 Table 32. Girls' estimated group probabilities versus the proportion of the sample assigned to the group ............................................................................................................................. 117 Table 33. Boys' estimated group probabilities versus the proportion of the sample assigned to the group ............................................................................................................................. 117 Table 34. Predicted average BMI for the girls............................................................................ 119 Table 35. Predicted average BMI for the boys ........................................................................... 120 Table 36. Classification of girls' BMI trajectories ...................................................................... 121 Table 37. Classification of boys' BMI trajectories ..................................................................... 121 Table 38. Descriptive statistics for the girls' model.................................................................... 122 Table 39. Descriptive statistics for the boys' models.................................................................. 124  xii Table 40. Distribution of parents’ (surrogates’) age, income, and education and association with girls’ BMI trajectory group membership .................................................................... 126 Table 41. Distribution of parents’ (surrogates’) age, income, and education and association with boys’ BMI trajectory group membership.................................................................... 128 Table 42. Mean parenting practices and temperament scores of girls with different BMI trajectories........................................................................................................................... 130 Table 43. Mean parenting practices and temperament scores of boys with different BMI trajectories........................................................................................................................... 132 Table 44. Results of discriminant function analysis: Structure matrix (girls) ............................ 136 Table 45. Results of discriminant function analysis: Classification results (girls)..................... 136 Table 46. Results of discriminant function analysis: Structure matrix (boys)............................ 138 Table 47. Results of discriminant function analysis: Classification results (boys) .................... 139               xiii List of Figures Figure 1. Overview of the model (adapted from Nagin, 1999) .................................................... 72 Figure 2. Proc Traj output: Girls' 4-group model with all-cubic trajectories.............................. 105 Figure 3. Proc Traj output: Boys’ 3-group model with all-cubic trajectories............................. 107 Figure 4. Trajectories of mean BMI and 95% confidence intervals (girls) ................................ 114 Figure 5. Trajectories of mean BMI and 95% confidence intervals (boys)................................ 116 Figure 6. BMI trajectories for girls ............................................................................................. 122 Figure 7. BMI trajectories for boys............................................................................................. 124 Figure 8. Canonical discriminant function (girls)....................................................................... 135 Figure 9. Canonical discriminant function (boys) ...................................................................... 138            xiv Acknowledgements My deepest gratitude is to my supervisor Dr. Pam Ratner, for her guidance and council, and for having faith and confidence in me throughout my doctoral studies. I am especially grateful for all her patience, critical review of the work, and open mindedness, without which I could have not embarked on a topic of such an interdisciplinary nature. I am also grateful to my committee members for all their comments and valuable suggestions; Dr. Susan Dahinten, for her unique scholarly perspective, insightful feedback, and guidance, and Dr. Sheila Marshall for helping me refine my understanding of pertinent parenting concepts, which greatly assisted my investigation. I thank my parents, sisters, and brother, for their lasting support and love. To my parents, I am indebted for instilling in me the appreciation of knowledge and education, and I am grateful also to my mother-in-law and my late father-in-law for their confidence in me and for their support. I cannot end without thanking my family, on whose encouragement and love I have relied throughout my time in the program. Most of all, my deepest appreciation goes to my husband Hazem for his unwavering faith in me and for his love. His patience and support helped me to overcome many challenging situations and to finish this dissertation. Last but not least, I thank my children Dina and Faris, for their unconditional love: Dina, for your kindness and sensitivity, and Faris, for your sense of humorous mischief.   1  Chapter 1: Introduction The prevalence rate of overweight status among Canadian children between the ages of 7 and 13 years, in a 15-year-time span (from 1981 to 1996), has increased from 15.0% to 28.8% in boys, and from 15.0% to 23.6% in girls (Tremblay & Willms, 2000). Obesity, in the same time period, increased from 5.0% in both genders to 13.5% in boys and 11.8% in girls. Based on data from the third cycle of the Canadian National Longitudinal Survey of Children and Youth (NLSCY) (1998/1999), it was reported that, of children aged 2-11 years, 37% were overweight and 18% were obese (Statistics Canada, 2002).1 In the United States of America (USA), the prevalence of childhood obesity has tripled in a period of 25 years (Schwimmer, 2003). The progressive increase in the prevalence of overweight status and obesity among Canadian children and adolescents is more advanced than that observed in Britain and Spain (Tremblay, Katzmarzky, & Willms, 2002). In a study of English children, the percentage of children classified as overweight and obese increased from 14.7% and 5.4%, respectively, in 1989, to 23.6% and 9.2%, in 1998 (Bundred, Kitchiner, & Buchan, 2001). Alarmingly, the trend toward overweight status is being noticed in children as young as 3 years of age. Childhood obesity is typically defined as the accumulation of high body fat or adipose tissue in relation to lean body mass that adversely affects health (World Health Organization [WHO], 2000). Classification of the weight status of children (overweight or obese) corresponds to the probability of various health-related outcomes. Obesity among  1 An estimated 35% of girls and 38% of boys fell into the category “overweight,” which includes the 17% of girls and 19% of boys who were classified as obese (Statistics Canada, 2002).  2 children and adolescents is a serious and challenging health issue. It is one of the most prevalent and increasing health problems in both developed and underdeveloped countries (WHO, 2000). The trend toward obesity and overweight status among children and adolescents has increased among all age groups, races, income levels as well as in both genders (Troiano, Flegal, Kuczmarsky, Campbell, & Johnson, 1995). Overweight status, obesity, or high body mass index (BMI)2, as it is usually assessed, if present in childhood, typically continues into adulthood (Parsons, Power, Logan, & Summerbell, 1999; Whitaker, Wright, Pepe, Seidel, & Dietz, 1997) and progressively increases with age. Guo and Chumlea (1999) noted that there is an 80% risk of developing adult obesity at age 35 years in persons who were obese at nine years of age or older. Obesity is a health phenomenon that involves both the physical and psychosocial (emotional, social, psychological) well being of humans. Pediatric obesity adversely contributes to children’s short-and long-term physical and psychosocial health outcomes and is likely to contribute largely to adult-onset morbidity. The latent or cumulative effects of obesity in childhood result in higher adult morbidity and premature mortality. The complex associations between pediatric obesity and its health-related consequences have been investigated extensively in diverse perspectives.3 Regardless of perspective, however, it is readily evident that the seriousness of childhood obesity lies in its health, social and  2 Body mass index (BMI), used to define nutritional status, is derived from the formula: weight (kg)/height (m)². It is the most frequently used measure of weight status and is usually used to classify underweight, overweight and obesity in children, adolescents and adults; it can also be used to estimate the prevalence rate of obesity within a population and to identify the health vulnerabilities associated with obesity (WHO, 2000).  3 These diverse perspectives include the study of the association between obesity and various health outcomes (e.g., physical, cognitive, and psychological), trends and changes in prevalence among gender, cultural, and socioeconomic groups, contributing and risk factors, and the management and treatment of obesity.  3 economic consequences; children with an unhealthful body weight are at risk for developing complex physical, cognitive, emotional and psychosocial health problems. It is associated with pulmonary, endocrine and metabolic disorders, and cardiovascular, musculoskeletal, and liver conditions (Mustillo et al., 2003; Reilly et al., 2003; Schwimmer, Burwinkle, & Varni, 2003). The increase in the prevalence of obesity in children and adolescents raises concern about the accompanied increase of other chronic diseases, some of which manifest during childhood. Overweight and obese children may develop asthma, Type II diabetes mellitus, stress on their joints, sleep apnea, and high blood pressure. Obese children are at increased risk for cardiovascular diseases (Must, 1996); indeed, the process underlying coronary heart disease has been found to start in children as young as three years of age (Newman et al., 1986). Childhood obesity has been associated with a 1.5 times increased risk for all cause mortality and a 2-fold increased risk of cardiovascular mortality during adulthood (Must & Strauss, 1999). Additionally, children who are overweight are more likely to become overweight adults (Dietz, 1998); a childhood diagnosis of obesity increases the relative risk of adult obesity by a factor of 5.3 in boys and 6.7 in girls (Kotani et al., 1997). Another underestimated and under diagnosed impact of obesity on children’s well being is one that is not mediated through physical illness, but through a reflection of social attitudes about body weight and body aesthetics. Society’s attention to body weight has children ‘socialized’ to the importance of appearances early in life (Jonides, Buschbacher, & Barlow, 2002) resulting in children that are socially responsive to slimness at a young age (Dietz, 1998) and a view of obesity as a cosmetic rather than a health issue (Schwimmer, 2003). Research results exhibit disturbing trends; children in general prefer to be friends with children with physical challenges rather than with obese children (Richardson, Goodman,  4 Hastorf, & Dornbusch, 1961). School-aged children identify thin people as having more friends, and being better looking, smarter and neater (Harris & Smith, 1983). Feldman, Feldman, and Goodman (1988) found children as young as 5 and 6 years old expressing distaste for obese bodies and being overweight as something negative. Recent research findings report that children and adolescents who are obese are more likely to experience impaired health-related quality of life when compared to healthy children and adolescents (Schwimmer et al., 2003). Obese children and adolescents are teased at school (Schwimmer et al., 2003) and are more likely to be the victims of bullying (Janssen, Craig, Boyce, & Pickett, 2004). Studies confirm that the most common consequences of obesity in children are psychosocial (Dietz, 1998; Reilly et al., 2003) manifesting in poor self esteem (Ravens- Sieberer, Redegeld, & Bullinger, 2001), negative self image (Davison & Birch, 2001),4 depression (Friedlander, Larkin, Rosen, Palermo, & Redline, 2003), isolation (Dietz, 1998), poor social adjustment, and poor school performance, achievement and attendance (Laitinen, Power, Sovio, & Jarvelin, 2002; Mo-Suwan, Lebel, Puetpaiboon, & Junjana, 1999). Children who are obese have early onset physical maturation, specifically in height (Dietz, 1998),5 which places them at greater risk for early and systematic discrimination and stigmatization by teachers and peers (Gortmaker, Must, Perrin, Sobol, & Dietz, 1993). Regrettably, obese children often are negatively viewed and stereotyped by health-care professionals, even those  4 Children with problematic weight and who are as young as 5 years of age develop negative self image (Davison & Birch, 2001).  5 Dietz (1998) documented that obese children tend to develop faster (especially in height) compared to children with normal weight and that the difference in body size rendered obese children more prone to discrimination.   5 who are ‘expert’ in obesity treatment (Teachman & Brownell, 2001). As a result of the increasing health and social consequences of obesity, a parallel increase in the associated health-care costs is to be anticipated. In Canada, the estimated costs associated with obesity were 1.8 billion Canadian dollars or 2.4% of overall health-care spending in 1997 (Birmingham, Muller, Palepu, Spinelli, & Anis, 1999). Using more recent internationally-based obesity definitions, the economic burden of obesity in 2001 accounted for 4.3 billion Canadian dollars (1.6 billion for direct costs and 2.7 billion for indirect costs) (Katzmarzky & Janssen, 2004).6 In the last few decades, exploration of the negative connotations of obesity has resulted in growing awareness of the short term consequences of obesity in children and adolescents. Many researchers (using longitudinal and cross sectional designs) have focused on the psychopathological and psychosocial aspects (Mustillo et al., 2003), the socioeconomic consequences (Laitinen et al., 2002), or the physical health problems (Freedman, Dietz, Srinivasan, & Berenson, 1999) arising from obesity. All their studies have shown significant associations between obesity and adverse health-related outcomes. These studies are obviously valuable but they are limited, particularly because they have not explored factors that may contribute to the development of obesity. Researchers have not yet provided a comprehensive understanding of the development of obesity in children. Yet this understanding will significantly contribute to identification of solutions to the problem including strategies for primary and secondary prevention.  6 The authors also studied the economic impact of physical inactivity, which accounted for 5.3 billion Canadian dollars.   6 Hence, to further our understanding of the developmental processes in the variability of childhood obesity, the primary purpose of this study was to identify, in a large representative sample of Canadian children, the age-related trajectories of overweight and obesity from toddlerhood into school age. To prevent childhood obesity, there is a need to focus on the identification and understanding of risk factors that lead to problematic weight status. In addition to the well documented risk factors associated with the high prevalence of childhood obesity (i.e., diet, exercise, and sedentary behavior), it also has been reported that childhood obesity is associated with the family environment and with children’s intrinsic characteristics. Researchers have documented relationships between parenting and children’s BMI and between children’s temperament and their BMI. Only one study has combined childhood temperament and parenting. To date, there are no published works that have considered the combined effect of these two factors and the development of obesity; moreover, no published work has examined the relationships among temperament, parenting, their interaction, and BMI in a nationally representative sample of Canadian children. Therefore, this study was designed to investigate the relationship between temperament and parenting practices, within the context of the theoretical goodness-of-fit model, and to determine their capacity to predict the development of overweight and obesity in young children.    7 Chapter 2: Literature review and research questions This chapter is divided into three main sections: (a) a review of the documented risk factors associated with childhood obesity, (b) a review of the pertinent literature about the relationships between childhood obesity, temperament and parenting practices, and (c) a description of the principal research questions. Risk factors associated with childhood obesity The published literature has documented that childhood obesity is influenced by complex combinations of genetic, environmental, socio-cultural, and behavioral factors (Betz, 2000).7 Associated or contributing factors for weight problems in children include ethnicity or race, lower socio-economic status, gender, single-parent households, high caloric consumption, and an imbalance between dietary intake and energy expenditure, increased fat cells, the presence of health or psychosocial disorders, children’s dietary and activity practices and attitudes, lack of physical activity, engagement in sedentary activities, parental overweight status or obesity, parenting practices, attitudes and beliefs, and having harsh and abusive parents (Agras, Hammer, McNicholas, & Kraemer, 2004; Buiten & Metzger, 2000; Costello, Keeler, & Angold, 2001; Davison & Birch, 2001; Felitti et al., 1998; Fontvieille, Harper, Ferraro, Spraul, & Ravussin, 1993; Gill, 1997; Klesges, Klesges, Eck, & Shelton,1995; O’Loughlin, Paradis, Meshefedjian, & Gray-Donald, 2000; Reilly et al., 2003; Robinson, 1999; Schumann, Nichols, & Livingston, 2002; Sothern & Gordon, 2003; Stunkard & Sorensen, 1993; Taras, Sallis, Nader, & Nelson, 1990; Tremblay & Willms, 2003).  7 Unfortunately, researchers have tended to study these factors in isolation from one another.  8 Genetics and early life experiences Although some authors believe that environmental factors contribute as much as 80% to the causes of obesity (Sothern & Gordon, 2003), twin studies suggest that 25% to 50% of the tendency towards obesity is inherited (Kiess, Müller, Kapellen, & Böttner, 2001). Children with two obese parents have a 50% chance of developing obesity in their life time, while children with one obese parent have one half that risk of developing obesity (Whitaker et al., 1997). Other parental influences associated with obesity in children are passed on from both genetics and shared environments, including those arising from parental activity and parental food practices and attitudes (Deheeger, Rolland-Cachera, & Fontvielle, 1997). Dietz and Gortmaker (2001) suggested that there are several critical periods for the development of obesity: during the periods of intrauterine life to early infancy, the preschool years, and adolescence. The intrauterine period extending to early infancy is critical because of maternal eating habits during pregnancy. Whitaker and Dietz (1998) proposed that maternal obesity increases the transfer of nutrients across the placenta and as a result induces lasting transformations in infants’ appetite and metabolic functioning.8 High BMI during infancy is usually attributed to birth weight, feeding methods, and the gender of the child (Davies, Day, & Lucas, 1991; Li, O’Conner, Buckley, & Specker, 1995; Wells, Cole, & Davies, 1996), and rapid growth in infancy is associated with increased BMI at age 6 years (Gunnarsdottir & Thorsdottir, 2003). Above average birth weight is associated with high BMI in infancy (Ebbeling, Pawlak, & Ludwig, 2002) and in children 5 to 7 years of age (Danielzik et al., 2004). Infants with high birth weight have shown an increased risk for Type  8 The Dutch Famine Study (Ravelli, Slein, & Susser, 1976) showed that under nutrition occurring at critical periods during fetal development can induce permanent physiological changes that may result in obesity.  9 II diabetes mellitus later in life (Eriksson, Forsen, Osmond, & Barker, 2003). On the other hand, being small for gestational age is associated with obesity in young children (Sothern & Gordon, 2003).9 Obese children less than 3 years of age have a lower risk of developing adulthood obesity if their parents are within normal weight limits compared with children with parents with problematic weight (Whitaker et al., 1997). Although infant weight has not fully predicted subsequent BMI, feeding methods during infancy have predicted weight at a later childhood age. Bottle fed infants are at higher risk of developing obesity later in life compared to breast fed infants (Gillman et al., 2001). The second critical period for the development of obesity is during the preschool years when ‘adiposity rebound’ occurs.10 Normally, body fatness decreases after the first year of life and increases between the ages of 5 and 8 years (Froidevoux, Schutz, Christine, & Jequier, 1993; Sothern & Gordon, 2003). Obesity in children over 6 years of age has been shown to be a strong predictor of adulthood obesity.11 Although adiposity rebound in children aged 5-6 years is associated with adulthood obesity (Whitaker et al., 1998), the most important factor contributing to obesity in children up to 6 years of age, however, is parental obesity (Cooper, Poage, Barstow, & Springer, 1990; Davies, Gregory, & White, 1995;  9 Some researchers have documented a relationship between low birth weight and central obesity in adulthood (Kensara et al., 2004).  10 Adiposity rebound is a point where body fatness increases after a period of decline to a minimum. It is usually related to adult obesity (Dietz, 2000).  11 Researchers have documented that childhood obesity (at less than 13 years of age) is moderately associated with adulthood overweight, and the prediction of adulthood overweight is excellent from the weight of individuals at 18 years of age (Guo et al., 1994). However, Power, Lake, and Cole (1997) found that the majority of their sample of individuals aged 33 years was of normal weight at ages 7, 11, and 16 years of age. The discrepancies in the findings of studies that have attempted to predict adult weight from childhood weight may be related to the application of different references or cutoff criteria for BMI in children.  10 Sothern & Gordon; Whitaker et al., 1997). Adolescence, a phase with strong emotional and social ramifications (Sanci, Glover, & Coffey, 2003), has been associated with changes in health behavior that may increase proneness toward overweight status and obesity. Indeed, adolescence may be the greatest period of risk for the development of obesity. During adolescence, puberty is accompanied by body fat changes such as alterations in fat distribution with female adolescents accumulating more fat than boys (Gazzangia & Burn, 1993). Adolescents are increasingly reported to: be more involved in sedentary activities, consume high caloric and high fat diets, be prone to eating disorders, and spend relatively little time engaged in physical activity (Biddle, Gorely, & Stensel, 2004; Guillaume, Lapidus, Becker, Lambert, & Bjorntorp, 1995).12 Physical inactivity Being overweight or obese in childhood has been frequently associated with an imbalance between dietary intake and energy expenditure (Sothern & Gordon, 2003; Tremblay & Willms, 2003),13 and physical activity is the most changeable element of energy expenditure (Wells et al., 1997). Consequently physical inactivity is theorized to contribute to the development or maintenance of childhood obesity (Baranowski et al., 1992; Tremblay & Willms, 2003; Trost et al., 1996). Sedentary behavior, such as watching television, playing computer games, and the absence of regular physical activity, is a documented risk factor for obesity (Gill, 1997; Schumann et al., 2002; Taras et al., 1990; Tremblay & Willms, 2003).  12 Female adolescents are more inactive than their male peers.  13 When children experience minimal or stable physical growth, the balance between their physical activity and food consumption predicts subsequent changes in their body weight (Klesges, Klesges, Eck, & Shelton, 1995).  11 Recently, children as young as 3 to 5 years of age have demonstrated limited involvement in physical activity (Strauss & Knight, 1999) and are more engaged in sedentary activities (Fontvieille et al., 1993; Reilly et al., 2003). However, infant physical activity is considerably variable among individuals (Wells & Davies, 1995) and no explicit relationship between infant activity and obesity has been established (Wells et al., 1996). Prevention of obesity Because of the complexity of the problem and the adverse consequences of childhood overweight and obesity, primary prevention is likely the most important intervention. Early action can prevent or minimize obesity and overweight and their associated psychosocial, cognitive and health outcomes (Bussey & Morgan, 1997). The treatment or management of obesity, once a child has become obese, is difficult (Barlow & Dietz, 1998) and rebound is expected.14 Rosenbaum, Leibel, and Hirsch (1998) noted that 90% to 95% of individuals who lose weight subsequently regain it. Furthermore, as children grow, it becomes more difficult for them to lose their excess weight.15 Unfortunately, little attention has been given to the primary prevention of childhood obesity both in practice and research; more effort has been placed on treatment that emphasizes changing behavior, dietary patterns, and levels of activity through education (Jonides et al., 2002; Schumann et al., 2002), which in itself is problematic because it tends to ignore the contextual factors that influence those behaviors. Researchers are now evaluating the effectiveness of programs for the prevention of obesity that include a range of lifestyle interventions (they typically focus on diet and physical  14 Rebound is the gaining back of weight lost.  15 The added difficulty observed in children’s treatment programs for obesity has been linked to many factors including peer pressure and psychological and cognitive immaturity (Ebbeling et al., 2002).  12 activity); alas, the results of these investigations indicate that programs that focus on changing the behavior of children also result in insignificant changes in the rates of obesity and overweight (Campbell, Waters, O’Meara, Kelly, & Summerbell, 2004; Warren, Henry, Lightowler, & Perwaiz, 2003). Summerbell et al. (2004) suggested that further research should consider the psychosocial determinants of behavior change. The majority of studies on obesity have been cross-sectional in design and thus little is known about the developmental processes in the variability of childhood obesity. Even less has been documented about the factors that accompany children’s growth and development, which will have to be studied longitudinally, and whether certain patterns of obesity are associated with particular factors. Temperament and obesity Although research attention has been paid to children’s behavior as a correlate of overweight and obesity, surprisingly little attention has been paid to children’s temperament. Temperament is broadly defined as individual differences in behavioral tendencies “in emotional, motor, and attentional reactivity and self-regulation” (Rothbart & Bates, 1998, p. 109). The behavioral manifestations of temperament are visible during infancy and are relatively stable with development. While recognizing its biological foundation,16 it is pertinent to acknowledge that the nature and manifestation of temperament is constantly modified by interactions with the environment (Sanson & Rothbart, 1995). The possible role  16 Observable behavior of temperament is assumed to be biologically based (i.e., behavioral manifestations of temperament are biological in nature; such as in the regulation of sleep and eating habits, psychological processes that include cardiac reactivity (Rothbart & Bates, 1998). The notion that temperament is biologically founded is supported by the fact that the dimensions are found in different cultures and the strong evidence of the heritability of the dimensions (Costa & McCrae, 2001)  13 of temperament in the development of childhood obesity has been recognized for a long time; nonetheless, few researchers have documented the extent to which temperament contributes to childhood obesity. The relevance of temperament for research on overweight and obesity could become apparent, considering that temperament is associated with an increased vulnerability to problematic eating (Martin et al., 2000) and poor health habits, such as sedentary activity (Räikkönen & Keltikangas-Järvinen, 1991). Mehrabian and Riccioni (1986) found that a predisposition to obesity, uncontrollable urges to eat, and a predisposition to anorexia are all correlated positively, albeit weakly, with some measures of temperament. The published research on obesity and temperament has been mostly conducted in the context of the metabolic syndrome (Keltikangas-Järvinen, Räikkönen, & Solakivi, 1990; Räikkönen, Matthews, & Salomon, 2003; Ravaja & Keltikangas-Järvinen, 1995; Ravaja, Keltikangas-Järvinen, & Keskivaara, 1996). Metabolic syndrome is a constellation of clinical findings including hyperinsulinemia, hyperglycemia, an increased plasma concentration of very low density lipoprotein (VLDL) triglycerides, a decreased plasma concentration of high density lipoproteins (HDL), obesity, and hypertension (Ravaja & Keltikangas-Järvinen, 1995; Viner, Segal, Lichtarowicz-Krynska, & Hindmarsh, 2005).17 The coexistence of these somatic parameters predicts cardiovascular morbidity and mortality. The risk factors contributing to the metabolic syndrome have been documented in early childhood and are found to continue into adulthood (Berenson et al., 1998; Katzmarzky et al., 2001), including weight gain and obesity in childhood (Bavdekar et al., 1999), excessive caloric intake, and sedentary lifestyle (Ferguson et al., 1999). What is of interest here is that there is evidence  17 Much of the literature points to insulin resistance as the underlying defect of the metabolic syndrome (DeFronzo & Ferrannini, 1991) and a consequence of obesity (Karam, 1992; Viner et al., 2005).  14 that temperamental dispositions may contribute to the development of metabolic syndrome in both adults (Räikkönen et al., 2003) and children (Ravaja & Keltikangas-Järvinen, 1995; Ravaja et al., 1996). Researchers have described relationships between several indices of hostility and behavioral and physical risk factors associated with the metabolic syndrome in children (Wills, Cleary, Shiner, & Fegan, 2002).18 For example, Keltikangas-Järvinen et al. (1990) provided evidence of a link between dimensions of temperament (specifically, the degree of hostility or physical activity)19 and physiological variables including body mass index, diastolic and systolic blood pressure, and serum lipid levels in both children and adolescents. In longitudinal studies, high activity temperament and aggression were documented to predict higher levels of factors comprising the metabolic syndrome (Ravaja & Keltikangas-Järvinen, 1995; Ravaja et al., 1996). More specifically, a relationship was found between negative emotionality (aggression and anger) and high BMI in 15-year-old boys. Evidence of an association between temperament and high BMI is provided by a few, limited research studies. One of the earliest studies, by Carey, Hegvik, and McDevitt (1988), established a significant relationship between child temperament and both rapid weight gain and obesity in middle childhood. In their longitudinal study, these researchers reported that temperament characteristics, including low adaptability, high emotional intensity, and withdrawal were predictive of rapid weight gain between the ages of 3 and 7 years. More  18 Hostility is a complex behavior or personality dimension that is correlated with emotional negativity in infancy and childhood; negative emotionality represents a predisposition for angry and aggressive behavior (Cairns & Cairns, 1991; Ledingham, 1991).  19 Examples of indicators of temperament specific to activity, used in these studies, included: eating, walking, and rapid speech. Researchers have reported a relationship between rapid eating and obesity among infants, preschoolers and school-age children (Agras, Kraemer, Berkowitz, Korner, & Hammer, 1987; He Ding, Fong, & Karlberg, 2000; Marston, London, & Cooper, 1976).  15 specifically, Carey (1985) reported that infants that gained the most weight were described by their mothers as ‘difficult;’ the most prominent temperament characteristic related to rapid weight gain was negative and intense mood.20 However in both studies (Carey, 1985; Carey et al., 1988), obesity and overweight were determined by plotting the research participants’ weight for height percentiles, using the US National Center for Health Statistics growth charts, rather than the more recent internationally-based cutoff points for obesity and overweight status that utilize BMI centile charts. More recently, Wells et al. (1997) reported a significant relationship between temperament assessed at 12 weeks and an increased risk for overweight measured until 2-3.5 years. In this prospective study, infants who showed distress to limitations (i.e., waiting for food, being refused food, or being confined) had greater percentages of body fat while easily soothable infants had lower skinfold thickness values. Of the 50 children in the study cohort, 20 were lost to follow-up, and the infants were all described as Caucasian and with mothers who reported high levels of education. Another recent longitudinal, prospective study revealed a significant correlation between high BMI (assessed at 9.5 years of age) and highly emotional temperaments (measured as an active personality with a predilection for anger or frustration) (Agras et al., 2004).21 The study cohort (n = 216) was not representative of the general population and their temperament was measured at age 5 years.  20 Rapid weight gain or growth in the first few months of life is a risk factor for overweight status in children (Stettler, Zemel, Kumanyika, & Stallings, 2002).  21 To date, this is the only study that has investigated multiple risk factors including both children’s factors (i.e., temperament, infant weight status, infant feeding behavior, early weight gain, childhood eating behavior, 24-hour child caloric intake, and child activity) and parents’ factors (parenting styles, maternal expectations of infant feeding and habits, parent/infant feeding practices, parents’ eating behavior, parents’ concerns about the child’s weight, maternal weight gain, maternal return to work, and parents’ weight status).   16 Anderson, Bandini, Dietz, and Must (2004) found that girls aged 8 to 12 years with relatively high activity temperament were leaner than girls with low activity temperament.22 Based on the results of their follow-up study, however, Anderson, Bandini, and Must (2005) failed to confirm that high activity temperament is a protective factor against increased adiposity (body fat) at four years after menarche. Moreover, Kremer et al. (1986) failed to find relationships between infant temperament and overweight in a prospective study of 351 healthy infants followed from birth until age 24 months. The measures of temperament were collected at an extremely early age, and the sample was described as being from diverse socioeconomic backgrounds. The major studies in the field all possess significant limitations in large part because they have relied on relatively small samples recruited mainly from private American clinical practices that provide services to middle class families or parents with relatively high education, who are not representative of the general population; thus the studies have limited generalizability. The pathways between temperament and obesity, a promising hypothesis, have been explored in a limited fashion. Researchers have speculated that the link with obesity, if it exists at all, occurs because of a tendency to overfeed children with difficult temperaments (Carey, 1985). Temperament, however, should be viewed as an ‘open system’ (Rothbart & Bates, 1998), where “new environmental features can emerge independently or can result from the child-parent interactional process. The same process can modify or change abilities,  22 Body composition estimated by total body water was used to identify obesity. The term "body composition" refers to the constituents that provide for structure, movement, and metabolic functions within the body. Body fat is one component of body composition and is often expressed as a percentage of total body weight. Body composition can not be measured directly, thus all methods of body composition measurement are only estimations.  17 motives, behavioral style or psychodynamic defenses, this process continues at all age periods” (Chess & Thomas, 1999, p. 89). The effects of temperament on weight status, if causal in nature, likely occur in interaction with several other variables. Associations between any particular variable and children’s weight status are best explored in the presence of interactions of other determinants that affect children’s weight. Given these premises, temperament is best studied longitudinally and in conjunction with assessments of the children’s environments, especially those aspects created by their parents.  A central aspect of children’s environments, which has great significance for children’s outcomes, is the parental influences to which they are exposed. In the literature reviewed on the relationship between temperament and obesity, only one study that focused on established and hypothesized risk factors included parental behavior. In that study, children’s temperament was assessed at one time (at 5 years of age), and parental behavior was shown to be statistically non-significant in predicting obesity and overweight at age 9.5 years (Agras et al., 2004). Parenting and obesity The health status of children is complex and multidimensional in nature. The physical and social sustenance of children is not contingent on their constitutions alone, but also on their families’ beliefs, values and behavior as well as on their socioeconomic status and education. Parents are the primary source of children’s cognitive, physical, and affective wellbeing, which is embedded in larger political, psychosocial and economic contexts.23 The  23 Studies have documented that parenting is interrelated with family environmental factors. Among such factors are single parent status, low maternal education, young maternal age, and parental psychological wellbeing. Further, parenting practices may differ among different socio-cultural backgrounds and contexts.  18 centrality of the family unit establishes patterns of health practice and behavior conducive to the development of the child and provides a strong argument for evaluating parenting as an important element of children’s health (Tinsley, 2003). To further develop this field of research, the link between temperament and high BMI is explicated by exploring a critical factor in the development of childhood obesity: parenting.24 Parenting is described as the “set of behaviors involved across life in the relations among organisms who are usually conspecifics…whose interactions provide resources across the generational groups and function in regard to domains of survival, reproduction, nurturance, and socialization” (Lerner, Castellino, Terry, Villarruel, & McKinney, 1995, p. 285). Parents are responsible for the creation and maintenance of an environment that is favorable to children’s growth and development; parents establish, guide and enforce children’s health behavior, including their ability to control and regulate their own health behavior (Barber, 1996; Tinsley, 2003). That is, parents instill in their children a sense of agency in their own life (Grolnick, 2003). The influences of parenting on childhood obesity are attributed to both genetic (parental weight) and multiple environmental factors and the interactions between the genetic component and environmental contexts. The role of the familial environment in the study of childhood obesity has received greater research consideration in recent years. Predictors of childhood obesity that have been related to the home environment and thereafter to obesity include food selection and preference, parental dietary intake, activity patterns, home eating patterns, meal structure, nutritional knowledge (level of education), cognitive stimulation,   24 Family context is important to every other domain of children’s development and for that reason, the assessment of parenting is important in the identification of risk factors for obesity.  19 and family functioning (Birch & Fisher, 1995, 2000; Chen & Kennedy, 2004; Gable & Lutz, 2000; Johnson & Birch, 1994; Johnson, Birch, & McPhee, 1991; Kinston, Loader, & Miller, 1987; Kinston, Loader, Miller, & Rein, 1988; Lissau-Lund & Sфrensen, 1992, 1994; Nicklas, Baranowski, Cullen, Rittenberry, & Olvera, 2001; Strauss & Knight, 1999; Tinsley, 1992; Wilkins et al., 1998). Researchers have documented relationships between parenting styles and problematic eating behavior (Field et al., 2001; Hill & Franklin, 1998; Humphrey, 1989; Russell, Kopec- Schrader, Rey, & Beumont, 1992). Of particular importance to this work, researchers have shown that there are negative relationships between authoritarian and permissive styles of parenting and the general wellbeing of children (Tinsley, 1992). For example, home environments that lack control and are described as anarchic have been associated with bulimia in adolescent girls (Agras et al., 1999), while strict and controlling parenting practices have been associated with anorexia in girls (Hill & Franklin, 1998; Russell et al., 1992). Researchers also have shown negative relationships between certain dimensions of parental feeding practices and general parenting styles and children’s weight status. It is recognized that parents can influence their children’s health behavior in several ways; Davison and Birch (2001) reviewed potential pathways through which parents shape their children’s nutritional behavior and potentially contribute to their weight status. These pathways include parental dietary knowledge, role modeling of eating behavior, through the provision of healthful environments, and parental feeding styles of their children. According to Wardle (2003), specific feeding styles that could influence the development of childhood obesity include: (a) family eating patterns (shared meal times or watching television at meal  20 times), (b) parental control over children’s eating whether in the use of restriction, monitoring or pressuring, (c) emotional feeding as in cases of offering food to regulate temper tantrums, and (d) instrumental feeding such as offering food as a reward or withholding it as a punishment. Christoffel and Forsyth (1989) suggested that the development of early and severe obesity may be related to certain patterns of family dysfunction (i.e., parenting practices). They attributed the development of obesity to the inability of parents to deal with excessive demands for food or children’s needs for different forms of attention and stimulation. Similarly, Story et al. (2002) reported a lack of family involvement as a barrier that interfered with treatment efforts for children with weight problems. Fisher, Mitchell, Smiciklas-Wright, and Birch (2002) found that when parents pressured their children to eat fruits and vegetables there was relatively lower consumption of fruits and vegetables among 5-year-old children. Other studies about feeding styles have documented that some parents use food as a strategy to control emotional outbursts (e.g., temper tantrums) and these feeding practices are associated with obesity in both the child and mother (Sherman et al., 1995). Similarly, Birch and Fisher (1996) reported that parents used food as a reward or withheld it as a form of punishment, which resulted in a reduced inclination to consume healthful foods. These two feeding practices are theorized to lead the child to associate eating as a response to cues other than hunger, thereby increasing the risk of high BMI. However, Wardle et al. (2002), Baughcum et al. (2001), and Sallis et al. (1995) failed to replicate these findings. In Wardle et al.’s (2002) study, emotional and instrumental feeding styles did not predict obesity in children although parents’ emotional feeding and children’s emotional eating were  21 significantly correlated.25 One of the most studied feeding practices is restrictive feeding. Restrictive feeding practices are those in which a parent limits the amount or type of food consumed by a child or controls the time of feeding (Birch, Fisher, & Davison, 2003). Evidence of the influence of restrictive-feeding practices on body weight has been equivocal and the impact of these strategies appears to be gender specific. Nonetheless, there is consensus that restrictive feeding can result in overeating and is associated with childhood overweight (Birch et al., 2003); it is recognized, however, that, in some cases, childhood overweight may contribute to parents adopting restrictive feeding practices (Fisher & Birch, 1999). Some researchers have documented a negative correlation between parental restrictions and eating habits and BMI, while other researchers have failed to report a relationship, and some have even reported a positive relationship. For example, Fisher and Birch (1999) documented that girls tend to prefer foods that parents restrict and when access to those foods is given, girls consume more of them. Longitudinal research has provided further evidence to support this notion about girls. Birch, Fisher, and Davison (2003) reported that 5-year old girls exposed to more restrictive feeding practices tend to eat more food, in the absence of hunger, than they do at ages 7 and 9 years. Further, Robinson, Kiernan, Matheson, and Haydel (2001) found that parental control and restriction over food intake are inversely associated with overweight (measured by BMI) in third grade girls, but not in boys. Birch and Davison (2001) and Johnson, Birch, and McPhee (1991), on the other hand, showed that greater maternal restriction or control during mealtime was associated with children having relatively more  25 Please refer to Faith et al. (2004) for some thoughts about the reasons for these discrepancies in the empirical studies that focused on the relationships between a variety of feeding styles and the development of obesity.  22 body fat. In addition to the influence of restrictive feeding practices on children’s weight, researchers are showing that there is an emotional toll of restricting practices on 4-6-year-old girls. For example, Fisher and Birch (2000) reported that girls have feelings of guilt and shame when eating restricted snacks. Other researchers have failed to support the hypothesized relationship between feeding practices and later maladaptive eating habits, overweight, or obesity. Saelens, Ernst, and Epstein (2000) found no support for the hypothesis that parental control over feeding is related to obesity in children 7 to 12 years of age, while Faith et al. (2004) reported lower BMIs in children 5 years of age with parents who monitored their food intake. Furthermore, Spruijt-Metz et al. (2002) found that white and African American boys and girls with mothers who pressured them to eat had lower total body fat mass, once total lean mass and energy intake were taken into account. Kremers et al. (2003), who also studied the relationship between parenting and children’s eating behavior, suggested that parental control of children’s feeding practices may have constructive influences on the children’s outcomes if situated within an atmosphere of warmth rather than a more demanding and rigid environment. This notion is further supported by Lees and Tinsley (1998), who found that mothers who gave higher than average amounts of praise, kisses, and hugs, as rewards for good behavior, had children who were more independent in tooth brushing, hand washing, and exercising, and that these children were more likely to make nutritious food choices. Conversely, De Bourdeaudhuij and Van Oost (2000) found that adolescents who described their parents as indulgent had unhealthful dietary habits (e.g., they consumed more foods high in fat and sugar, and more snacks); further, these adolescents (who had more decision-making power concerning their food  23 selection) described fewer healthful food selections by their families. Although these studies all examined parenting styles in the context of feeding-related practices, they give an indication of the larger and more general styles that parents adopt. Kremers et al. (2003) stated that a ‘general parenting style’ could be viewed as “an environmental context factor that may influence the effectiveness of parental child-feeding practices” (p. 44). In addition to the documented relationship between feeding-related practices and eating habits and weight status in children, a growing body of empirical literature, albeit with conflicting conclusions, has explored the relationships between children’s weight and particular parenting practices and parenting styles. Felitti et al. (1998) reported a positive relationship between children’s exposure to harsh parental disciplinary styles (use of physical punishment) and adulthood obesity. Additionally, Christoffel and Forsyth (1989) noted that the parent-child interactions of children who are obese are characterized by ineffective limit-setting and neglect. In a cross- sectional study, a relatively democratic parenting style was found to be related to higher BMI in children 8 to 11 years of age (Chen & Kennedy, 2004). Lissau-Lund and Sфrensen (1992) reported that families of obese children have a tendency to be negligent or to show less compassion. This was further described in their prospective, population-based study of children who were 9 to 10 years of age and followed for 10 years (Lissau-Lund & Sфrensen, 1994). Parental support was found to have a highly significant effect on the risk of obesity in young adulthood. Further, overprotective support tended to predispose children to an increased risk of obesity, although the effect was not statistically significant. An additional study, worthy of mention, is the work of Strauss and Knight (1999) who reported an increased risk for the development of obesity in children living in environments with low  24 levels of cognitive stimulation (i.e., parent-centered activities indicating responsiveness), independent of socioeconomic factors, race, maternal BMI, and maternal marital status. On the other hand, Mustillo et al. (2003) found no associations between harsh or overprotective parenting, lax supervision, and the development of obesity. Strauss and Knight (1999) also reported no association between parental support and the development of obesity. In their study, children who became obese were just as likely to be hugged, kissed or spanked as were children who did not develop obesity. Further, Gable and Lutz (2000) were unable to predict obesity in children, 3 to 10 years of age, from their parents’ parenting styles (authoritative or authoritarian). Agras et al. (2004), in a longitudinal study that followed children from birth to 9.5 years of age, documented no associations between three parenting styles (i.e., authoritarian, authoritative, or permissive styles of behavior) and overweight and obesity.26 In addition to the impact of parenting practices and parents’ control of what, how and when children eat, parents are highly influential with respect to their children’s level of activity and sedentary behavior, and consequently their weight. In their study of the influence of parental feeding styles on children’s health behavior, Arredondo et al. (2006) reported that the children of parents that monitored and reinforced healthful behavior engaged in more physical activity, ate more healthful foods and ate fewer unhealthful foods. Gentile and Walsh (2002) reported that children watched less television and engaged in more alternative activities when their parents applied consistent discipline (e.g., putting limits on the amount of time television was watched). More distinctively, children 4-13 years of age, who participated in the first cycle of the Canadian National Survey of Children and Youth  26 It is important to note that, in their publication, it is not clear at what time the parenting style was assessed or the frequency at which the assessments were made.  25 (NLSCY, 1994), who resided with parents with consistent, non-punitive or non-aversive parenting styles, were shown to be involved in higher levels of physical activity (Cragg et al., 1999). Similarly, children residing with parents who endorsed authoritative parenting styles (i.e., highly demanding and responsive) were found to be more active than were children of parents with authoritarian parenting styles (i.e., highly demanding and low in responsiveness) (Gable & Lutz, 2000). It is established that the home environment and characteristics have an enormous impact on children’s health and well being, including the development of obesity. The modest literature on parenting practices and the development of obesity, however, has been inconsistent and any conclusions drawn are debatable. Most studies have been cross-sectional and cannot provide evidence of causality, or its direction. In the studies reviewed, the measures of parenting practices or styles were either focused on narrow aspects of parenting (e.g., affective expression) or were type-specific (e.g., parenting styles: authoritarian versus authoritative). Additionally, the studies that relied on type-specific measures were based on computed total scores for the parenting styles typology, rather than on specific observations of overt behavior. This unfortunately does not allow for a distinction to be made between parents who are supportive and attuned with their children and those who are overly attentive to, or overprotective of, their children, or between parents who are both demanding and warm and those who are rejecting and demanding. Gender differences in the impact of parenting practices on feeding behavior and weight status have been noted repeatedly; nonetheless, the literature does not elucidate or explain those differences.27 And, in the  27 Holm-Denoma et al. (2005) reported that parents differed in how they reported the feeding behavior of girls and boys. For example, mothers were more likely to report that their  26 literature reviewed, none of the studies that explored parenting and childhood obesity considered the contribution of an interaction between parenting and temperament. The discussed limitations of the various studies that have examined the relationship between parenting practices and childhood obesity suggest that this is an important area to be investigated, especially with longitudinal data, which may explicate the longitudinal effects of parenting on children’s BMI and well being. Parenting and temperament Although much thought has been given to the importance of parental behavior and how it influences children’s outcomes, there has been increasing awareness of the child’s own contribution to the parent-child interaction and the influence it has on various outcomes (Rothbart & Bates, 1998; Sanson et al., 2002). The majority of the work on the relationships between children’s characteristics and parental behavior and practices has focused on children’s temperament (Rothbart & Bates, 1998). In early conceptions of children’s socialization, developmental outcomes were perceived to be the result of the environment. In this unidirectional approach, parenting practices had direct effects on children’s outcomes. Measures of children’s outcomes were regarded separately and parental practices were believed to be directly responsible for those effects (Schaffer, 1999) with no acknowledgment of any influence arising from a child’s constitutional nature or temperament. Thomas and Chess (1977) observed individual differences in infants’ primary patterns of reaction, based on their personal experiences, which were subsequently investigated in the New York Longitudinal Study (NYLS) (Thomas, Chess, & Birch, 1968).  daughters ate enough food and that they had good appetites in comparison with their sons’ feeding behavior.  27 Results of their study showed that children contributed to their own development, which contradicted earlier beliefs about the unidirectional influences of the environment. The outcomes of development were found to be influenced and affected by characteristics specific to the child. After the results of the NYLS were acknowledged, researchers started focusing on the inherent temperamental characteristics that exerted their effects on development, with little attention paid to the environment.28 This unidirectional perspective resulted in studies that adopted correlational methodologies and with results that demonstrated that temperament had direct linear effects on parent-child relationships (Sanson, Hemphill, & Smart, 2004; Scarr & McCarthy, 1983). The unidirectional approach to the study of children’s development has produced most of the published literature and is based on correlations between children’s development and individual differences. Researchers have interpreted these correlations as evidence of the unidirectional influence of temperament on children’s outcomes (Sanson et al., 2004). A second model that explains the processes by which temperament affects development explicates the indirect effects of individual differences on development (Bates, Dodge, Pettit, & Ridge, 1998; Sanson et al., 2004). This model posits that conditions of the socializing environment, such as parenting practices and socioeconomic conditions have significant influences on the expression of temperament characteristics. This perspective encompasses an approach where temperament alone does not predict development; it does so in concurrence with particular environments (Bates et al., 1998; Mangelsdorf, Gunner,  28 Factors contributing to development are not exclusively related to temperament, other prominent contributors are sex, gender, physical attributes and infant prematurity. For more details on the subject, refer to Schaffer (1999).  28 Kestenbaum, Lang, & Andreas, 1990). More recently, researchers have focused on the goodness-of-fit theoretical model, which considers the relationship between children’s temperaments and their environments. The model could best be described as an ‘interactionist’ developmental model where “behavioral phenomena are considered to be the expression of a continuous organism- environment interaction from their very first manifestations in the life of the individual’ (Chess & Thomas, 1999, p. 13). The goodness-of-fit model, proposed by Thomas and Chess (1977), specifies that children’s development is affected by the consonance (goodness-of-fit) of the child’s characteristics and his or her environment and its expectations and demands (Chess & Thomas, 1999). Goodness-of-fit is embedded within “the values and demands of a given culture or socioeconomic group” (Thomas & Chess, 1977, p. 112). Furthermore, good fit occurs when the child’s characteristics are adequate to cope with the expectations of the environment in which the child functions, thereby enhancing healthful development. Poor fit refers to an incompatibility between the child’s behavioral style (i.e., temperament) and parental expectations, abilities and circumstances; a child’s behavioral characteristics may have an intrusive or disruptive effect on his or her environment. Poor fit may increase the child’s vulnerability to maladaptive functioning and problematic development (Chess & Thomas, 1996, 1999). In the goodness-of-fit model, parenting is viewed as dynamic, interactive states wherein children’s temperaments affect their parents’ parenting practices and thus bring forth different responses. Developmental psychologists emphasize that these interactions are reciprocal---adults influence their children’s behavior and children influence their parents’  29 behavior and handling of the children.29 This was described by Chess and Thomas, as early as 1977, when they postulated that the influence of the infant’s temperamental traits is determined by the opportunities, constraints and demands of family members, teachers, and peers. In parallel processing, the influence of the family and social community is determined by the quality and degree of its consonance or dissonance with the child’s abilities and style of functioning.30 Depending on the degree of the goodness-of-fit, children with similar initial characteristics may end up with very different behavioral styles. For example, in a consonant environment, a child with a difficult temperament may not develop problematic behaviors. Temperamental characteristics that are consonant with the environment will allow children to deal positively with problematic situations thus having better control over their emotions; control over negative emotional responses produce better problem-solving abilities (Eisenberg et al., 2000; Rothbart & Ahadi, 1994) and consequently positive development. As Thomas and Chess (1977) stated, individual children relate to their uniquely individual contexts to produce uniquely individual outcomes. The application of the goodness-of-fit model implies knowledge of specific interactions between the child and the parent. For example, it is pertinent to assess the environmental contexts within which the child exhibits the manifestations of his or her temperament (affect, distractibility or withdrawal). Additionally, parental reactions cannot be adequately assessed without concurrent consideration of the child’s temperamental characteristics and their influence on the parent (Chess & Thomas, 1999).  29 Perceived social support is also crucial. Research is showing that parental behavior and mental well being and child adjustment also are influenced by the level of social support one receives.  30 They also stressed that the outcomes of temperament need to be considered mutually within the larger context of the child’s environment.  30 Based on clinical observations, Thomas and Chess (1977) concluded that children described with a ‘difficult’ temperament are more vulnerable to a variety of behavioral problems if their parents are inconsistent, impatient or pressuring.31 Using the same data, Cameron (1977) concluded that parenting practices that are rejecting and inconsistent in the use of discipline are related to negative alterations in children’s temperament. Further, the study identified a difference in the temperamental changes observed among boys and girls when the parent applied strict practices. Parental strictness was associated with decreased adaptability and negative mood in boys and with less persistence in girls. In an attempt to explore the theoretical concept of ‘goodness-of- fit’, Gordon (1981) divided children according to their parents’ rating of their temperament - into an “easy” and “difficult” group - and instructed the parents of 3-year olds to use either a highly structured and controlling parenting practice or to employ a relaxed, permissive practice. The study conclusion was not statistically significant although suggestive of a poor fit between boys’ difficult temperaments and controlling parents (described by the author as authoritarian parenting practices) when compared with permissive parenting practices. Conversely, girls with difficult temperaments responded favorably to the controlling practices. The observed impact  31 Thomas and Chess (1977) identified three general dimensions which group children according to their temperament profiles; they are the “easy child”, “difficult child” and the “slow to warm up child.” In these constellations, difficult children usually experience negative moods (loud periods of crying), display intense responses, have irregular biological patterns, withdraw, are slower in adapting and frustration produces negative and intense reactions. Children rated ‘easy’ are adaptable to change, accept most frustrations with little fuss, readily accept rules, have regular biological rhythms, are in relatively good moods most of the time and are attracted to novelty. ‘Slow to warm up’ or shy children tend to withdraw from novelty and adapt to change slowly, their emotional reactions are usually negative but of low intensity (mild intensity in both negative and positive reactions). More recently, researchers have endorsed three broad dimensions of temperament: reactivity or negative emotionality, self-regulation, and approach-withdrawal/inhibition or sociability (Sanson et al., 2004).  31 of controlling parenting practices on girls contradicts Cameron’s findings. Lee and Bates (1985) investigated the impact of a variety of maternal-child interactions, in 2-year-old girls, that were grouped according to whether they exhibited manifestations of a difficult temperament or an easy temperament. They attempted to understand the mediating process (between a difficult temperament and behavioral problems) by examining how these interactions related to the perceived temperamental difficulties. The authors concluded that children with difficult temperaments experienced more conflictive interactions with their mothers’ control attempts.32 Furthermore, these mothers tended to apply more intrusive parenting practices than did mothers with children of easy-average temperaments (these mothers used more nurturing types of practice, positive interactions and control). Within the context of childhood obesity, the role of temperament is best studied in a framework of parenting practices; parents provide the environments that are detrimental to their children’s development. In the relevant literature uncovered, only one study by Agras et al. (2004) combined both temperament and parenting practices in an attempt to explain obesity in children. They found a statistically non-significant association between parenting practices and weight status and thus removed the variable from all their multivariate analyses. This statistically non-significant result, however, could be attributed to the late assessment of the children’s temperament (5 years of age). It is well documented that temperament is best assessed at an earlier age so as to avoid the confounding effects of the environment.  32 The child’s reactions included engaging in troublesome behavior, and the child became more negative or resistant to her mother’s control. The mother, on her part, continued to use more intrusive control practices. The authors concluded that these child-parent conflictive interactions placed the children at higher risk for problematic behavior due to the vicious cycle of poor-fit encounters between the child’s disposition and the parental parenting practice.  32 Potential pathways to obesity Little is known about how children’s characteristics may contribute to childhood obesity; genetics plays a significant role in determining both temperament and weight status, and they are both highly conditioned by environmental processes. Unraveling these processes is complex and is in need of further longitudinal and multidisciplinary studies. One potential approach for understanding these relationships is through understanding the process that may explicate how temperament characteristics interacting with parenting practices translate into behaviors that may eventually place a child at risk for the development of obesity. In the following sections, I describe untested but potentially explanatory theories that link parenting practices and children’s temperament to obesity. Temperament and maternal feeding practices The relationship between the effects of infants’ temperament and the ensuing patterns of maternal care has been investigated in the literature and the results are inconsistent. In their study of the maternal care of “difficult” children, some researchers have supported the notion that these children tend to elicit stronger maternal responses and attention (which is moderated by maternal education) (Crockenberg & Smith, 1982; Fish & Crockenberg, 1981), while others have concluded the opposite (Campbell, 1979; Lee & Bates, 1985). One important aspect of maternal care that is of interest to this field of research is feeding practices. According to Wardle et al. (2002), specific feeding styles that could influence the development of childhood obesity include emotional feeding, as in cases of offering food to regulate temper tantrums, and instrumental feeding, such as offering food as a reward or withholding it as a punishment or for comfort (Birch, 1998). Studies about feeding styles  33 have documented that some parents use food as a strategy to control emotional outbursts (e.g., temper tantrums) and that these feeding practices are associated with obesity in both the child and the mother (Baughcum et al., 2001; Jain et al., 2001). Maternal control of infant feeding may persist into childhood and parents who demonstrate the most control over their children’s food intake tend to have children who have difficulty adjusting their food intake in different settings (Johnson & Birch, 1994).33 Researchers have additionally established a relationship between having a difficult temperament and maternal feeding practices. In a study that investigated the relationship between the survival of infants living in severe famine areas and the infants’ temperaments, de Vries (1984) established that fussy, unadaptable, and intense infants were fed more and therefore survived famine, whereas the quiet and undemanding infants had higher prevalence of mortality. Replicating these findings in healthy infants, Wells et al. (1997) reported that infants with high soothability temperaments had less skinfold thickness. This finding prompted Carey (1985), who studied the relationship between difficult temperaments and rapid weight gain, to propose that infants with difficult temperaments gained more weight because their mothers fed the fussy infants more to calm them down. More recently, Agras et al. (2004) reported that persistent tantrums over food during childhood posed a risk for the development of obesity. Interestingly, these tantrums were correlated with some attributes of a highly emotional temperament. One additional proposed explanation of the association between temperament and obesity lies in maternal responses to difficult infants with intrusive feeding styles. It is documented that infants who are resistant to their mothers’ attempts at establishing regular  33 Klesges and Hanson (1988) suggested that parenting practices that promote internalization of values (authoritative) may lead to more self-control and therefore a reduced risk of obesity.  34 feeding patterns are evaluated as being difficult (Wolkind & de Salis, 1982). And, according to Schaffer (1996), parents who perceive any resistance to their care may see it as a form of rejection. Schaffer further suggested that these parents may, in some cases, continue their attempts of care, despite the distress it produces in the children. Consistent with Schaffer’s perspective, a mother who persists in her attempts to impose a feeding schedule on her ‘difficult’ infant may include forced feeding, which may contribute to the development of problematic eating habits that continue into childhood and possibly adolescence if intervention does not occur. Birch and Fisher (1998) concluded that parental lack of sensitivity to feeding cues, and strict control in feeding practices, can disrupt children’s ability to self regulate and may have adverse effects on children’s food preferences and intake (i.e., preference for high-fat foods, aversion to a variety of foods, and altering responsiveness to internal cues of hunger and satiety). To date, no longitudinal research has been conducted to identify correlations between infant feeding styles, later children’s eating styles, and obesity in the context of children’s temperament. Potential confounders: Parents’ education, age and income Potential confounding variables that could be associated with the development of childhood obesity and parenting are the age of the primary care giver, the household’s income and level of education. It is well documented that children growing up with educated mothers and residing in families with higher income levels are less vulnerable to a wide range of unfavorable behavioral and health outcomes. Parents’ educational attainment could affect their children’s weight status by influencing their knowledge of healthful nutrition and the need for physical and cognitive activities. Canadian studies have shown that educated parents tend to make healthier food choices and are more involved in physical activities with  35 their children (Tremblay & Willms, 2003). Children raised in families with lower educational attainment watch more television, on average (Bernard-Bonnin et al., 2002; Kimm et al., 1996), and are less likely to be cognitively engaged (Strauss & Knight, 1999), compared with their peers in families of higher educational attainment. All are factors related to an increased risk of childhood obesity and may be associated with the parenting practices of interest here. The prevalence of childhood obesity has reached epidemic proportions in children residing in families with lower socioeconomic status (SES), although studies that have explored the relationship between income and childhood obesity are inconsistent. This is attributed to different obesity definitions and indicators of SES having been employed (Sobal & Stunkard, 1989). Nonetheless, many studies in Canada and the US have documented an inverse relationship between income and obesity in preschool children (Johnson-Down, O’Loughlin, Koski, & Gray-Donald, 1997; Mei et al., 1998; Willms, Tremblay, & Katzmarzky, 2003) and in school-aged children (Wang, 2001; Wang, Monteiro, & Popkin, 2002). Further, compared with children residing in higher income families, children in lower income families consume less fruit, watch more television (Casey et al., 2001), and consume food with high fat content. The third potential confounding demographic variable that may affect the weight of children and parents’ parenting practices is the age of the parent. Research in child development has demonstrated a negative relationship between maternal age and a variety of adverse outcomes for children (Dahinten & Willms, 2002). For example, teenage mothers (< 20 years) tend to apply less responsive and more punitive child-rearing practices (Fergusson & Woodward, 1999). Mothers in the 25-29 year age range show less rejection of their 11- to 13-year-old children compared with mothers in the 13-24 year age range (Dahinten, Shapka,  36 & Willms, 2007). In their study of the relationships between measures of temperament and familial and children’s characteristics, Jabel, Normand, Tremblay, and Willms (2002) found that teenage mothers tend to rate their children lower on good-natured, independent, consistent, adaptable, and obedient temperamental scales. It is thus assumed that younger mothers are at higher risk of having children who are obese or overweight through the relationships between age, educational attainment and income, and through the effects of age on parenting practices. Summary of the research problem Several studies have pointed to the effects of various social contexts on children’s BMI. Most of the investigations of factors that contribute to high BMI have been limited to environmental conditions and circumstances without consideration of children’s personal characteristics that may interact with their environments and make them prone to obesity. The relationships between obesity and temperament have been explored in a limited fashion, and not in the context of parenting. The prevention of obesity is considered a global health priority; if we are to accomplish this goal, we must explore new ways of thinking. Researchers have documented relationships between parenting and BMI and between BMI and temperament. Oddly, only one study has combined child temperament and parenting. To date, there are no published findings that have considered the combined effect of these two variables and the development of obesity; moreover, no published work has examined the relationships among temperament, parenting, their interaction, and BMI in a nationally representative sample of Canadian children. Furthermore, few studies have been longitudinal in design and thus the potential to describe different clusters of obesity, and its development, with distinct contributing factors has been limited.  37 There is great need to study a representative sample of Canadian children to provide a robust and generalizable description of the development of obesity and to identify key risk factors. Therefore, the purpose of this study was to identify, in a large representative sample of Canadian children, the age-related trajectories of overweight and obesity from infancy into childhood. The intent was to investigate the associations between these trajectories and children’s temperaments, their parents’ parenting practices and their interactions. Potential confounders (i.e., parental age, income level and educational attainment) were recognized as important. The study was designed to integrate what is known about individual factors (child temperament) with knowledge of social contextual factors, at the parental level (parenting practices) to elaborate our understanding of the mechanisms that lead to high BMI. The goal was to understand the contributing factors of high BMI through the study of the reciprocal individual-environmental factors that affect weight status in children. This work is based on the assumption that many individual characteristics (i.e., health behavior) are highly contingent on social processes (i.e., socio-ecological contexts) and that individual and social level variables mutually shape health and disease (Diez-Roux, 1998; Evans, Barer, & Marmor, 1994). It is believed that environmental conditions interact with the characteristics of the individual to produce variations in individual health status. For example, variations may arise from compositional effects with children with particular natures being more likely to become obese because of individual characteristics found in particular home settings. This notion of reciprocal individual-environmental interactions complements the concept of goodness-of-fit developed by Thomas and Chess (1977). Thus, the application of the goodness-of-fit concept provides a suitable developmental framework for the study of temperament and childhood obesity.  38 Research purpose This study was designed to investigate the relationship between temperament and parenting practices, within the context of the goodness-of-fit model, and to determine their capacity to predict the development of overweight and obesity in young children. The goals of the study were: (a) to determine the number and type of distinct trajectories in the development of overweight and obesity in a representative sample of Canadian children who were between 24 and 35 months of age, at baseline, and followed for 6 years (using Cycles 2 through 5 of the Canadian National Longitudinal Survey on Children and Youth (NLSCY) data; (b) to investigate the associations between children’s temperament characteristics and their parents’ parenting practices, and the clusters of trajectory in BMI change over time; and (c) to investigate the interactive contributions of specific factors of temperament and parenting practices to the trajectory clusters of BMI. Research questions The following research questions were posed: • What are the number and types of distinct trajectories in the development of obesity in a representative sample of Canadian children who are between 24 to 35 months of age, at baseline, and followed biennially over a 6-year span? • What is the relationship between children’s temperament and the development and maintenance of obesity (BMI trajectory group membership)? o Is there a difference in the development and maintenance of obesity (BMI trajectory group membership) between children with “fussy/difficult” and “not fussy/difficult” temperaments?  39 • What is the relationship between different parenting practices and the development and maintenance of obesity (BMI trajectory group membership)? o Are different parenting practices associated with different trajectories of change in BMI over time? • Are temperament and parenting practices independent predictive factors or synergistic in the development of childhood obesity?       40 Chapter 3: Methods Data source and sample The Canadian National Longitudinal Survey of Children and Youth (NLSCY) is a wide-ranging and ongoing survey conducted through a partnership between Statistics Canada and Human Resources and Social Development Canada (HRSDC). The survey is designed to produce both longitudinal and cross-sectional estimates to monitor changes in the development and well being of children and to assess determinants of children’s health within their environments. The collection of a ‘national database’ of the characteristics of children’s biological, social and economic contexts, is anticipated to allow health policy makers to develop programs that are needs-oriented, innovative and conducive to the well being of Canadian children.   The sample size of the NLSCY Cycle 1, conducted in 1994, consisted of approximately 23,000 children ranging in age from newborn to 11 years; the survey design was based on cluster sampling. The starting point for the NLSCY sample was the household. Participating households were mainly selected from Statistics Canada’s Labour Force Survey (Main Component) and from the National Population Health Survey (Integrated Component) (both based on the Labour Force sample frame, which is representative of the Canadian population).34 The sampling frame excluded children living in institutions and children living on Indian reserves. The child, and not the dwelling, is the statistical unit. Once the household was selected, one child, 0 to 11 years of age, living the majority of the time in the selected household, was selected at random. Then, other children in the same economic family were  34 The Labour Force Survey uses a stratified, multistage probability sample design that is based on an area frame in which dwellings are the sampling units.  41 selected at random for a maximum of four children per household. In cases of twins, both children were selected. Sampling for the NLSCY was such that it is possible to produce both cross-sectional and longitudinal estimates. The NLSCY sample for Cycle 1 was allocated by both age group and then by province. Sample allocation was based on sufficient samples in each of seven key age groupings or cohorts: 0 to 11 months, 1, 2 to 3, 4 to 5, 6 to 7, 8 to 9, and 10 to 11 years of age. A sufficient sample was required to produce reliable estimates at the national level to reliably measure characteristics with a prevalence of 4% for each age group after five survey cycles. Further, a sufficient sample was obtained in each of the 10 provinces and territories to produce reliable estimates for all children aged newborn to 11 years in Cycle 1. Less than 1- to 11-month old and one-year old children were over sampled by keeping them in separate age groups. In Cycle 1, there were 22,831 responding children 0 to 11 years of age in 13,439 households. After sub-sampling, only 16,903 respondents from the longitudinal cohort are to be followed until they reach age 25 years.  In Cycle 2, two populations were targeted: the longitudinal sample representing the population of children aged 0 to 11 years living in a province in 1994, and a cross-sectional sample that covered children aged 0 to 13 years living in a province in 1996. For financial reasons, there was a need to reduce the size of the Cycle 2 sample. Consequently, certain Cycle 1 responding households were removed from the Cycle 2 sample, and of the households remaining, up to two longitudinal children were surveyed. Of 10,216 households retained (n = 20,025 children), 15,468 longitudinal children from the original cohort were  42 surveyed in Cycle 2.35 Follow-up of the longitudinal children who had changed their residence was accomplished by interviewing them in their new homes if they had relocated to one of the provinces, the Yukon, the Northwest Territories, the continental United States, to an Indian reserve, a military barracks, or an institution. Children who left North America were excluded from the follow-up surveys. A total of 38,035 children were sampled in Cycle 3. Only 3% of those sampled were out of scope and were not included. Three cohorts for the longitudinal sample were described in this cycle. Cohort one included the same initial sample as Cycle 2 with the exceptions of deceased children, duplicate cases, households that moved permanently out of the country, children on Indian reserves, and the non-respondents of the initial sample of Cycle 2. The longitudinal sample for the first cohort of Cycle 3 culminated in a total of 16,718 children. The second cohort included 2,506 children selected from the Labour Force Survey and 1,483 children who were the siblings of the Cycle 1 longitudinal children. A total number of 7,944 children made up the third cohort of Cycle 3. The sample included children from the Labour Force Survey and the Birth Register. To meet the analytical goals of describing certain aspects of children’s development, an additional 7,052 5-year old children were added to Cycle 3.  A total sample size of 36,789 children comprised Cycle 4. The cross-sectional population of Cycle 4 represented children who were 0 to 17 years of age on January 1, 2001. The longitudinal population of Cycle 4 consisted of 4 cohorts. Cohort one, represented all children who were 0-11 years old in Cycle 1, including 15,632 children from the longitudinal  35 Children aged 0 to 1 years were added to the sample to ensure cross-sectional representation. Further, a sample of children aged 2 to 5 years old, residing in New Brunswick, was included.  43 cohort of Cycle 1. This cohort was between 6-17 years of age in Cycle 4. A total of 1,086 children had dropped out of the longitudinal cohort sample from Cycle 1. The exclusion criteria for the first cohort of Cycle 4 included households that were non-responsive after two consecutive cycles, hard refusals, deceased children, those who had permanently relocated out of the country, and children who had not responded in Cycle 2 and had moved in Cycle 3. Cohort two represented all children who were 0-1 years old in Cycle 2, including a total of 2,632 from the Labour Force Survey and 978 siblings of Cycle 1 longitudinal children. Cohort two children were between the ages of 4-5 years in Cycle 4. A total of 6,383 children were selected from the Birth Register and 1,735 children from the Labour Force Survey were contacted again for the third cohort of Cycle 4. The third cohort represented all children who were between 0-1 years of age in the third cycle and who were between 2-3 years of age in Cycle 4. The last cohort included children aged 0 and 1 years from the Labour Force Survey from 5,031 unique households.  At Cycle 5, the sample was comprised of about 30,800 children and youth. The cross- sectional sample of Cycle 5 consisted of children aged 0-5 years on December 31, 2002. This sample consisted of 0-1 year olds selected in 2002, the 0-1 year olds selected in 2000, and the 0-1 year olds selected in 1998. The longitudinal sample was made up of three cohorts. Cohort one consisted of children aged 0-11 years at Cycle 1 and who were between 8-19 years at Cycle 5 (n = 12,523). The second cohort consisted of 0-1 year old children at the time of their selection in Cycle 3. This cohort was between 4-5 years of age at Cycle 5. The final longitudinal cohort was comprised of children who were between 0-1 years at the time of their selection in Cycle 4 and were between 2-3 years at Cycle 5. Table 1 presents a summary of the number of respondents to the surveys, the number  44 of children included in the longitudinal cohort nested within the larger cross-sectional surveys (the group of interest here), and the response rates for each of the five cycles. The longitudinal cohort consisted of the children who were sampled in Cycle 1 at 0-11 years of age and who will be followed until they are 25 years of age. Table 1. Summary of the samples included in each survey cycle   Cycle 1 Cycle 2 Cycle 3 Cycle 4 Cycle 5 Number of respondents (longitudinal & cross-sectional combined) 22,831 20,025 38,035 36,789 30,800  Response rate 83.6% 92.8% 88.2% 84.8% 80.6%  Longitudinal cohort only  16,903 15,468 16,718 15,632 12,523   Study cohort The study cohort for this study was drawn from the NLSCY, Cycles 2 to 5 datasets. Children between the ages of 24 to 35 months in Cycle 2 of the NLSCY were selected as the baseline. In the subsequent cycles, only children with data in the second cycle (i.e., baseline) were selected for the longitudinal file. Data collection Data collection for the NLSCY, initiated in 1994, is conducted at two-year intervals. At this time, six cycles have been collected and the first five have been released through the  45 Research Data Centre Program.36 Data were collected from the person most knowledgeable (PMK) about the child. The PMK was asked to answer three sets of questionnaires. The first, the General Questionnaire, contained questions about both the PMK and his or her spouse or partner related to socio-economic issues, such as their education, labor force activity and income. The second set was the Parent Questionnaire that involved gathering information about the child’s social environment (e.g., the PMK’s mental health, social support, family functioning and characteristics of the environment).37 The third set, the Child’s Questionnaire, was completed for children aged 0 to 11 years. For this questionnaire, the PMK responded to questions related to parenting and child care and the child’s health, behavior, education, and literacy. The interviewer also administered the Peabody Picture Vocabulary Test - Revised (PPVT-R) to children aged 4-5 years to measure their receptive vocabulary. Teachers and principals were asked about the child’s achievement and behavior, and children over 10 years of age completed questionnaires related to their family, friends, school, and teachers and about their behavior, feelings, possible depression symptoms, self esteem, and health-related behavior. Children in grades two or above also responded to cognitive measures in the questionnaire. The data were collected by way of a face-to face or telephone interview using computer assisted interviewing (CAI), with the exception of the 10-11 year olds’  36 The Research Data Centre (RDC) program is part of an initiative by Statistics Canada, the Social Sciences and Humanities Research Council and university consortia to provide researchers with access to data from surveys. Access is provided to researchers with approved projects who have been sworn in under the Statistics Act as 'deemed employees.' Guidelines for access to the longitudinal data of the NLSCY are found in: www.statcan.ca/english/rdc/application.htm (see Appendix A)  37 The General and the Parent Questionnaires were combined to form one survey instrument for Cycles 2 and on.  46 Questionnaire. The interviewer also completed a questionnaire about the neighborhood of the participant, as a supplement to the information provided by the PMK, and a second questionnaire related to the conditions under which the PPVT-R test was administered. Written parental consent was obtained granting permission to obtain the relevant data from the schools and permitting the children to complete the 10-11 year olds’ Questionnaire, if applicable, and the PPVT-R test. Measures BMI: Obesity and overweight Conceptual definition: BMI  Obesity is defined as an excess accumulation of body fat (WHO) while overweight refers to having excess body weight in relation to height (Bellizzi & Dietz, 1999). Because the direct measurement of body fat is difficult and costly (WHO), anthropometric measures such as height and weight, skinfold thickness, and body mass index (BMI) have been the more conventional methods of determining weight status. Among these measures, BMI38 has been used most frequently and has been shown to have relatively strong correlations with body fat and with increased morbidity and mortality in adults (Gallagher et al., 1996; Troiano, Frongillo, Sobal, & Levitsky, 1996). Pietrobelli et al. (1998) confirmed that correlations of BMI with percent body fat ranged from 0.79 to 0.83 in 5- to 19-year-old girls and boys.39 Moreover, BMI has been described as a ‘consistent measure’ of obesity in  38 BMI is body weight (kg) divided by height squared (m²). BMI correlates with height, and to adjust for the differences in height through growth, the square power is used (Bellizzi & Dietz, 1999).  39 Depending on the sample age, gender, and ethnicity, the correlation coefficient (between BMI and body fat) ranges from 0.39 to 0.90 (Barlow & Dietz, 1998).  47 children across age groups (Barlow & Dietz, 1998) and thus has been recommended by the International Task Force on Obesity to provide a reliable index of body fatness (Dietz & Robinson, 1998). Although BMI is an objective measure, there is little consensus about how to classify or categorize children by BMI. In adults, the identification of overweight and obesity is done by empirically decided fixed cutoffs; in children, weight status is typically evaluated by ‘statistical definitions’ based on reference populations (Ogden, 2004). The statistical definitions, however, limit researchers’ ability to calculate the risk of existing or impending morbidity. Debates on how to define obesity and overweight in children arise, to a certain extent, because of the application of unsystematic centile cutoffs of selected reference data (Tremblay & Willms, 2000). The lack of a standardized classification system makes international comparisons difficult, if not impossible (Cole, Bellizzi, Flegal, & Dietz, 2000) and reaching an agreement on a criterion for the upper limit of normality in children is problematic (Dietz & Robinson, 1998). Three different sets of reference body mass index values are the most widely used in the assessment of overweight and obesity: (a) one developed by Must, Dallal, and Dietz (1991) based on the first US National Health and Nutrition Examination Survey; (b) Cole et al.’s (2000) values, which are internationally based;40 and (c) the growth charts for the US representing US national survey data of the Centers for Disease Control and Prevention   40 The cutoffs points were identified by pooling data on BMI for children from six different countries: Brazil, Great Britain, Hong Kong, the Netherlands, Singapore, and the US.   48 (Kuczmarski et al., 2000; Ogden et al., 2001).41 The World Health Organization recently developed new international standards of children’s growth that were released in 2006. These new standards have replaced the normalized WHO curves (Grummer-Strawn, Garza, & Johnson, 2002; WHO, 2006).42 Among children, BMI varies considerably with age, and gender differences in body composition also have been documented. This variability necessitates age- and gender- specific definitional criteria (Troiano & Flegal, 1999). Consequently, for this study, BMI was used as the reference method to measure overweight and obesity in children with specific cut-offs based on age and gender described by Cole et al. (2000). The classification, described below, is linked to adult health-related definitions of overweight and obesity. This approach has been used in other studies of Canadian childhood obesity including some that have been based on data from the NLSCY (cf., Katzmarzky & Janssen, 2004; Tremblay et al., 2002; Willms, Tremblay, & Katzmarzky, 2003). These cut-offs, developed by Cole et al. (2000),43 were based on pooled international data for BMI. The cutoffs are defined to pass  41 The criteria and definitions for overweight and obesity in children using the CDC BMI reference points (for age) have been endorsed by the Dietitians of Canada, the Canadian Pediatric Society, the College of Family Physicians of Canada, and the Community Health Nurses Association of Canada (2004). In this classification, BMI above the 95th percentile is defined as overweight, while BMI between the 85th and 95th percentiles is defined as ‘at risk for overweight.’ In their definition, the term ‘obesity’ is not used.  42 The World Health Organization Growth Standards were issued too late to be used in the analyses described in this dissertation; they are also limited, in terms of the work presented here, because they can only be applied to children up to 5 years of age. The new guidelines are based on growth patterns that should occur when infants are breastfed; it is believed that children worldwide should grow in similar patterns when they have been given good nutrition and healthcare. 43 Cole et al. (2000) identified centiles by pooling the results of corresponding overweight and obese 6 to 18-year-old boys and girls from six different population samples. The authors constructed centile curves for BMI for each data set (by sex) using the least mean squares method.  49 through BMI of 25 and 30 kg/m² at age 18 years, if the growth trajectory is maintained. The charts are gender and age specific and can be used for children 2 to18 years of age in six month intervals. Operational definitions: BMI, overweight and obesity  For the purpose of this study, BMI, overweight and obesity were operationally defined as follows: 1. BMI was defined as weight in kilograms over the square of the height in meters (kg/m²) (Statistics Canada, 2002; WHO, 2000). From the NLSCY datasets, BMI was calculated on the basis of the height and weight of the child reported by the ‘person most knowledgeable’ (PMK) about the child. This information was available for children 24 to 35 months of age, who participated in the second cycle of the NLSCY and for each cycle into the fifth cycle. Consequently, height and weight information was available at four different times (every two years). In the dataset, height was reported (without shoes) in feet and inches or in meters and centimeters, and weight was reported in kilograms and grams or in pounds and ounces. 2. The children’s reported BMI was compared with the international cutoff values for BMI for overweight and obesity, by sex, between the ages of 24-35 months in Cycle 2 through to 99-107 months in Cycle 5. These cutoff points (presented in Table 2), used to classify the children as being of normal weight, overweight, or obese, were adapted from Cole et al. (2000).    50 Table 2. Cole's obesity and overweight cut-offs*  Overweight Obese BMI Age (years) Boys Girls Boys Girls 2.0 18.41 18.02 20.09 19.81 2.5 18.13 17.76 19.80 19.55 4.0 17.55 17.28 19.29 19.15 4.5 17.47 17.19 19.26 19.12 6.0 17.55 17.34 19.78 19.65 6.5 17.71 17.53 20.23 20.08 8.0 18.44 18.35 21.60 21.57 8.5 18.76 18.69 22.17 22.18 9.0 19.10 19.07 22.77 22.81 * The cutoffs are defined to pass through BMI of 25 and 30 kg/m² at age 18 years, respectively, if the growth trajectory is maintained. Temperament Conceptual definition  Diversity in theoretical views and conceptualizations of temperament has contributed to many debates about the nature of temperament and has resulted in the development of several different measures (Bates, 1989; Rothbart & Bates, 1998).44 Notwithstanding these  44 The idea of childhood temperament has resulted in considerable discourse among researchers, clinicians and care givers. The controversy first started in the 1960's with the New York Longitudinal Study (NYLS) that relied on nine dimensions of temperament  51 debates, children’s temperament is generally regarded as the predominantly typical ways in which children experience and react to their environments. Temperament is the constitutionally-based and enduring individual difference in behavioral style (Rothbart & Bates, 1998; Sanson et al., 2004): in motor activity (Buss & Plomin, 1984), emotional and attentional reactivity (Goldsmith & Campos, 1986; Strelau, 1985; Thomas & Chess, 1977), and in patterns of self-regulation (i.e., self-soothing & persistence) (Rothbart, 1989; Rothbart & Derryberry, 1981; Sanson et al., 2002).45 Behavioral styles of temperament represent the ‘how’ of behavior and are distinguished from motivation and ability (representing the why and what of behavior, respectively) (Thomas & Chess, 1977).46 Dimensions of temperament constitute a subset of the more general area of personality; some authors consider temperament to be central to a child’s emerging personality. Researchers have related temperament to the Big Three and Big Five factors of personality (Rothbart & Bates, 1998;  (Thomas & Chess, 1977). Since then, many investigators have contributed to the development of the construct through debates about appropriate theoretical definitions, measurements, and the expectation of stability and consistency, and about linkages with social, physical and cognitive outcomes. Diversity in the theoretical views about childhood temperament has created miscommunication among researchers, practitioners and theorists. Among researchers, there remain uncertainties about what temperamental scales actually measure (i.e., there are uncertainties about the origins of temperament, the pertinent behavior, and whether it reflects constitutional factors or contextual influences). Further discussion of this issue is found in a chapter by Rothbart and Bates (1998).  45 For example, temperament represents the differences between two children who share the same age, developmental capabilities and motivations, but who exhibit differences in the regularity of their biological habits (sleeping and eating), their reactions to new schools, situations, and people, the ease with which they adapt to novelty, and their persistence in completing tasks.  46 Thomas and Chess (1977) viewed temperament as a behavioral style and thus any associated abilities, content or motivations were not included in their conceptualization of temperamental characteristics. Motivation in their view helps to shape behavior, which then has a motivational component (Chess & Thomas, 1989).   52 Wills et al., 2002). The term constitutional in the conceptual definition of temperament emphasizes its biological basis (Chess & Thomas, 1989), which is influenced by genetic inheritance (Buss & Plomin, 1984),47 maturation, and experience (Rothbart & Bates, 1998). Recent studies have identified some of the neurotransmitters and neurophysiological aspects of temperamental individuality (Bates & Pettit, 2007). Dimensions of temperament are characterized in terms of individual differences rather than universal patterns of development. Even though temperament reflects an increasingly and relatively consistent (over time), stable (across situations), and coherent pattern of behavioral style (Bates, 1989), the nature of expression and development of the characteristics of temperament (whether emphasizing regulatory and control components of behavior or the emotional expressiveness aspects of behavior (Rothbart & Derryberry, 1981; Strelau, 1985) can be constrained or enhanced by contextual factors. Children are obviously affected by developmental processes and the environments they exist within, and those environments operate on reactive individuals and not on ‘blank states of existence’ (Thomas & Chess, 1977). Various definitions of the “difficult” temperament have been utilized in studies, not withstanding these differences, the reported frequency and intensity of the negative effect are fundamental to the study of the relationship between temperament and children’s outcomes (Bates, 1980). For the purpose of this research, temperament was conceptually defined as individual differences in children’s behavioral styles with a focus on the fussy/difficult  47 Some aspects of temperament, such as activity, appear to be more strongly affected by heredity (Schmitz, Saudino, Plomin, Fulker, & DeFries, 1996). The genetic basis, however, does not indicate that temperament is fixed and unchangeable.  53 dimension. This includes the presence or absence of negative behavior and moods, such as difficulty in being soothed, irritability, fussiness, frustration, loud periods of crying, the making of happy sounds, and laughter, in addition to the intensity of those responses and behavior. In this conceptualization, if the child is described to be “not difficult or fussy,” then he or she would be observed to accept most frustrations with little fuss and to be in a relatively good mood, most of the time. On the other hand, a difficult/fussy temperament means that the child typically experiences negative moods (loud periods of crying, fussiness), displays intense responses, and frustration produces negative and intense reactions. The temperament scale The temperament scale used in the NLSCY was adapted from a well established scale originally known as the Infant Characteristics Questionnaire (ICQ). The ICQ was developed by Bates et al., in 1979, to measure parental perceptions of infant temperament with a focus on the difficult temperament.48 In the literature, the ICQ has been shown to be satisfactory in terms of its internal consistency and test-retest stability, is developmentally stable, and has demonstrated correlations with relevant behaviors (Bates & Bayles, 1984; Bates et al. cited in Lee & Bates, 1985; Finegan, Niccols, Zacher, & Hood, 1989; Olson, Bates, & Bayles, 1982). Furthermore, ICQ scores have been shown to possess convergent validity, there is a strong correlation between mothers’ and fathers’ ratings (Bates, Bennett-Freeland, & Lounsbury, 1979).  48 The items of the ICQ, originally suggested by Thomas and Chess, specified 9 dimensions of temperament including Prechtl’s changeability and soothability measures, and Robson and Moss’s fussiness and sociability measures (Bates et al., 1979). Mothers’ reports were found to be reliable over time (in particular, with the fussy-difficult dimension) and showed convergence with fathers’ reports; convergence has been also noted between ICQ factors and other comparable parent-report temperament instruments.  54 The ICQ, devised for 6-month old infants, is a short screening instrument usually based on mothers’ ratings. In the NLSCY, a revised version of the scale was used for children 0 to 3 years of age. In addition to the ICQ, the revised version of the temperament scale used in the NLSCY incorporated some items that were factor analyzed to support their inclusion. They are items from the Child Characteristics Questionnaire (CCQ) developed by Lee and Bates (1985) and the Preschool Characteristics Questionnaire (PCQ) developed by Finegan et al. (1989). The CCQ is a longer version of the ICQ and is used for 13- and 24- month old children, while the PCQ is an extension of the ICQ developed for preschool children. The age-specific items of the three scales, used in the NLSCY, permit the measure of temperament in children 3 to 47 months of age. The revised version of the ICQ, used in the NLSCY, collected data about the children that were between 1 and 3 years of age and theoretically measured the degree to which those children were irregular, unadaptable, persistent or unstoppable, and difficult. The person most knowledgeable (PMK), in most cases a parent, ranked the child’s difficulty in dealing with a described behavior on a 7-point scale. Scores ranged from 1, indicating that the child’s response was “favorable, or [the child] usually exhibit[ed] the specified behavior (rarely fussy)” to 7, indicating that the child reacted “negatively, or infrequently display[ed] the behavior” in question. A score of 4 represented ‘above average’. This scoring was applied to all questions with the exception of question TEMP-Q14, for which the interpretation of the score was reversed. In a factor analysis of the data collected in Cycle 1 of the NLSCY, only the difficult/fussy items factor loaded as expected; the factors ‘unadaptable,’ ‘dull’ and ‘unpredictable’ had item loadings that were inconclusive. As a result of these validation  55 problems in Cycle 1, the scale went through several alterations. In Cycle 2, only the ‘fussy/difficult’ construct was retained and in Cycle 3, the ‘unadaptable’ construct was reintroduced alongside the ‘fussy/difficult’ factor. In Cycle 4, the temperament questions were collected for children aged 0 to 2 years only and new items were added to the scale. Furthermore, the qualitative descriptor of the midpoint score of 4 was removed in Cycle 3 because the respondents found the response descriptor, ‘average’ to be confusing. Operational definition: Difficult/fussy temperament The construct temperament, adopted in this study, refers to constitutionally based individual differences in behavioral patterns (specifically emotional negativity) that were visible in the study cohort during Cycle 2 (when the children were aged 24 to 35 months). For this study, one dimension of temperament was adopted, the difficult/fussy construct (i.e., emotionality) initially described by Thomas and Chess (1977). The dimension has been consistently related to high BMI among both children and adults (Agras et al., 2004; Carey, 1985; Carey et al., 1988). The factor loadings of the items assessed during Cycle 1 provided more evidence of a factorial composition with Cronbach’s alpha = 0.78. Individual differences are affected by interactions between the child and his or her environment. In this research, only a child’s interactions with his or her parent, or surrogate, were investigated, although I acknowledge the contributions of the broader social context, and those of children’s motivations and cognitive development. It is important to acknowledge that the behavior style that a child demonstrates depends on the demands of the situation and is affected by the subjectivity of the reporter and the context.  56 Measures of fussy/difficult temperament The items from the NLSCY that were analyzed as indicators of the child’s ‘fussiness/difficulty’ dimension of his or her temperament are described in Table 3. One measure was obtained for each child from the Cycle 2 data. The items are age-specific thus there are 10 items for children between the ages of 24 and 35 months, the baseline age for the study cohort. Table 3. Age-specific measures of fussy/difficult temperament (24-35 months)   Variable Name NLSCY  Survey Question   Response Options BTMC-Q1 How easy or difficult is it for you to calm or sooth the baby? 1 ‘Very easy’ 2 3 4 ‘About average’ 5 6 7 ‘Difficult’ BTMC-Q5 How many times per day, on average does…get fussy and irritable—for either short or long periods of time? 1 ‘Never’ 2 ‘1-2 times/day’ 3 ‘3-4 times/day’ 4 ‘5-6 times/day’  57  Variable Name NLSCY  Survey Question   Response Options 5 ‘7-9 times/day’ 6 ‘10-14 times/day’ 7 ‘15 times/day or more’ BTMC-Q6 How much does he/she cry & fuss in general? 1 ‘Very little; much less than the average baby/child’ 4 ‘Average amount; about as much as the average baby/child’ 7 ‘A lot; much more than the average baby/child’ BTMC-Q7 How easily does he/she get upset? 1 ‘Very hard to upset-even by things that upset most babies/children’ 4 ‘About average’ 7 ‘Very easily upset by things that wouldn’t bother most babies/children’ BTMC- Q8A When he/she gets upset, how vigorously or loudly does he/she cry and fuss? 1 ‘Very mild intensity or loudness’ 4 ‘Moderate intensity or loudness’ 7 ‘Very loud or intense; rarely cuts loose’ BTMC- Q11 How much does he/she smile and make happy sounds? 1 ‘A great deal, much more than most infants/children’  58  Variable Name NLSCY  Survey Question   Response Options 4 ‘An average amount’ 7 ‘Very little, much less than most children’ BTMC- Q12 What kind of mood is he/she generally in? 1 ‘Very happy & cheerful’ 4 ‘Neither serious nor cheerful’ 7 ‘Serious’ BTMC- Q17 How changeable is …’s mood? 1 ‘Changes seldom & slowly when he/she does change’ 4 ‘About average’ 7 ‘Changes often & rapidly’ BTMC- Q19 On average, how much attention does he/she require, other than for caregiving (feeding, bathing, diaper change, etc.)? 1 ‘Very little-much less than the average child’ 4 ‘Average amount’ 7 ‘A lot-much more than the average child’ BTMC- Q33 Please rate the overall degree of difficulty… would present for the average parent. 1 ‘Very easy’ 4 ‘Ordinary, some problems’ 7 ‘Highly difficult to deal with’    59 Scoring The parent or the PMK answered the questions in the fussy/difficult temperament scale by choosing a rating between 1 and 7. With the responses summed, the lowest possible score of 10 meant that the child was low in reported fussiness and difficulty relative to his or her age group, while the highest possible score of 70 indicated that the child was highly fussy and difficult. Parenting practices Conceptual definition  Much of children’s development occurs within the family context, and parental interactions and involvement are critical to children’s psychosocial well being. Parenting is considered to be the most demanding and most complex responsibility of adulthood (Zigler, 1995). This responsibility entails feeding, protecting, stimulating, regulating affect, and enhancing social communication (Bornstein, 1995a). Parenting takes many forms about which few people concur; “variations in parenting philosophies, values, beliefs, ideas, and practices are widespread” (Bornstein, 1995b, p. xv).49 Notwithstanding this discourse, for the purposes of this study, I focused on parenting practices rather than on parenting styles.50  49 Further discourse on parenting involves the understanding of the process that leads from parenting practices to the child’s outcomes, the role of the child’s nature, the role of attachment, distinguishing between parental practices, behavior and attitudes, and what is considered to be optimal parenting. Much emphasis has been placed on the nature and dimensions of differences in parenting and on the conditions of parenting.  50 Darling and Steinberg (1993) suggested that a distinction be made between parenting styles (which are parental behaviors that encompass parenting practices) and other aspects of parent-child interaction; they involve emotional disposition, are not goal directed nor goal defined. A parent’s style is: (a) independent of the content and (b) is theoretically unrelated directly to a specific socialization event. They further stated that parenting practices can be  60 Parenting practices are the “behaviors defined by specific content and socialization goals” (Darling & Steinberg, 1993, p. 492) that include practices used by parents to help children reach goals associated with socialization. Consequently, for this study, parenting practices conducive to development were considered most relevant, including parental demonstrations of warmth through practicing positive interactions, consistency in discipline, and the lack of ineffective practices that would evoke aversion. The basis for the selection of these particular practices was the aim to locate maladaptive parental practices that could result in problematic weight status and, conversely, protective parenting practices that could promote healthful weight status. Parenting scales A set of 25 items in the NLSCY were used to measure parenting practices. The items, proposed and provided by Dr. M. Boyle for use in the NLSCY, were: (a) based on the work of Dr. K. Dodge and (b) an adaptation of Strayhorn and Weidman’s (1988) “Parent Practices Scale,” which examines the frequency of praise, approval of the child’s behavior, involvement, consistency in discipline, and hostility. Parents responded to the items with response options ranging from 1 to 5. Factor analyses were conducted by Statistics Canada to evaluate the psychometric properties of the items in the NLSCY population (Cycle 1); these factor analyses yielded four factors (for children 2-11 years): positive interaction, (Cronbach’s alpha = .81, k = 5), ineffective interactions (Cronbach’s alpha = .71, k = 7; item APRC-Q5 was not included for this age group), consistency in discipline (Cronbach’s alpha = 0.66, k = 5) and rational/non-aversional (Cronbach’s alpha = 0.57, k = 4). The factor  operationalized at different levels; depending on the outcome of interest (different parenting practices could be more relevant to some outcomes than to others).  61 loadings for some items for the rational factor were ‘insufficient to be included’ in subsequent analyses (i.e., APRC-Q19, -Q20, and -Q25) making a total of 21 items. Measures of parenting The items from the NLSCY that were analyzed as indicators of parenting practices are summarized in Tables 4, 5, and 6 (the scale totals to be used in this analysis were computed by Statistics Canada). Because of the low Cronbach’s alpha for the rational/non- aversional factor, it was excluded from the analyses. The dimension of positive interaction had 5 items; the dimension of consistency in discipline had 5 items, and the dimension of ineffective parenting had 7 items. Table 4. Positive interaction and scale rating  Variable Name NLSCY Survey Question  Response Options  BPR-Q1   How often do you praise.., by saying something like “good for you” or “what a nice thing you did” or “that’s a good thing”? 1 ‘ Never’ 2  ‘Once a week’ 3  ‘Few times a week’ 4  ‘One-two times/day’ 5  ‘Many times each day’ BPRC-Q2    How often do you and he/she talk, or play together, focusing attention on each other 5 minutes or more, just for fun? 1 ‘ Never’ 2  ‘Once a week’ 3  ‘Few times a week’ 4  ‘One-two times/day’  62 Variable Name NLSCY Survey Question  Response Options  5  ‘Many times each day’ BPRC-Q3    How often do you and he/she laugh together? 1 ‘ Never’ 2  ‘Once a week’ 3  ‘Few times a week’ 4  ‘One-two times/day’ 5  ‘Many times each day’ BPRC-Q6    How often do you do something special with him/her that he/she enjoys? 1 ‘ Never’ 2  ‘Once a week’ 3  ‘Few times a week’ 4  ‘One-two times/day’ 5  ‘Many times each day’ BPRC-Q7A  How often do you play games with him/her? 1 ‘ Never’ 2  ‘Once a week’ 3  ‘Few times a week’ 4  ‘One-two times/day’ 5  ‘Many times each day’      63  Table 5. Consistency in discipline and scale rating  Variable Name NLSCY Survey Question  Response Options  BPRC- Q10  When you give him/her a command or order to do something, what proportion of times do you make sure that he/she does it? 1  ‘Never’ 2  ‘Less than half time’ 3  ‘Half the time’ 4  ‘More than half the time’ 5  ‘All the time’ BPRC- Q11    If you tell him/her he/she will get punished if he/she doesn’t stop doing something, and he/she keeps doing it, how often will you punish him/her? 1  ‘Never’ 2  ‘Less than half time’ 3  ‘Half the time’ 4  ‘More than half the time’ 5  ‘All the time’ BPRC– Q12*    How often does he/she get away with things that you feel should have been punished? 1  ‘Never’ 2  ‘Less than half time’ 3  ‘Half the time’ 4  ‘More than half the time’ 5  ‘All the time’ BPRC- 16* How often is he/she able to get out of a punishment when he/she really sets his/her mind to it? 1  ‘Never’ 2  ‘Less than half time’ 3  ‘Half the time’  64 Variable Name NLSCY Survey Question  Response Options  4  ‘More than half the time’ 5  ‘All the time’ BPRC- 17* How often when you discipline her/him does he/she ignore the punishment? 1  ‘Never’ 2  ‘Less than half time’ 3  ‘Half the time’ 4  ‘More than half the time’ 5  ‘All the time’ *Item value reversed when total score calculated.   Table 6. Ineffective parenting and scale rating  Variable Name NLSCY Survey Question  Response Options  BPRC-Q4    How often do you get annoyed with….for saying or doing something he/she is not supposed to? 1 ‘Never’ 2 ‘Once a week’ 3 ‘Few times a week’ 4 ‘One-two times/day’ 5 ‘Many times each day’ BPRC-Q8* Of all the time you talk to …about his/her 1 ‘Never’  65 Variable Name NLSCY Survey Question  Response Options  behavior, what proportion is praise? 2 ‘Rarely’ 3 ‘Sometimes’ 4 ‘Often’ 5 ‘Always’ BPRC–Q9    Of all the times that you talk to him/her about his/her behavior, what proportion is disapproval? 1 ‘Never’ 2 ‘Less than half time’ 3 ‘Half the time’ 4 ‘More than half the time’ 5 ‘All the time’ BPRC–Q13    How often do you get angry when you punish….? 1 ‘Never’ 2 ‘Less than half time’ 3 ‘Half the time’ 4 ‘More than half the time’ 5 ‘All the time’ BPRC–Q14   How often do you think that the kind of punishment you give him/her depends on your mood? 1 ‘Never’ 2 ‘Rarely’ 3 ‘Sometimes’ 4 ‘Often’ 5 ‘Always’ BPRC–Q15 How often do you feel you are having 1 ‘Never’  66 Variable Name NLSCY Survey Question  Response Options    problems managing him/her in general? 2 ‘Rarely’ 3 ‘Sometimes’ 4 ‘Often’ 5 ‘Always’ BPRC–Q18  How often do you have to discipline him/her repeatedly for the same thing? 1 ‘Never’ 2 ‘Rarely’ 3 ‘Sometimes’ 4 ‘Often’ 5 ‘Always’ * Item value reversed when total score calculated.  Scoring A score for each measure was computed by totaling the values of the items making up the factor or dimension. To produce the total scores, the value 1 was subtracted from each item such that the total scores ranged from 0 to 4 times the number of items included in the factor. Thus, the total scores ranged between 0-20, 0-25, and 0-20 for the positive interaction, ineffective parenting and consistency variables, respectively. A score of zero on the factor demonstrated a lack of positive interaction; an absence of ineffective interactions for the ineffective factor; and an absence of consistency in discipline for the consistency factor.  67 Study design Data organization and analysis  The data analysis phase of the study proceeded in four stages. The first phase required merging the data from Cycles 2 through 5 into one data file. Second, the missing data were reviewed and descriptive statistics were obtained. Third, based on Nagin’s (2005) approach to group-based modeling, obesity trajectories were determined by fitting a semi-parametric mixture model (SPMM) to the data. This semi-parametric, group-based modeling strategy, explained in greater detail below, was used to identify the distinct groups of individual BMI trajectories over the ages of 24-35, 51-59, 75-83, and 99-107 months. This method allowed for the detection of different distinct classes of BMI change across ages, each with a specific intercept and slope and estimated population prevalence. For each of the trajectory groups, the model defined the shape of the trajectory and the estimated proportion of the population belonging to the trajectory group. Table 7 summarizes the age groups and variables measured for each of the survey cycles employed in the analyses to identify the BMI trajectories. Table 7. Age groups, survey cycles and variables measured  Age Survey Cycle Variable 24-35 months Cycle 2   Temperament Parenting BMI 51-59 months Cycle 3 BMI 75-83 months Cycle 4 BMI  68 Age Survey Cycle Variable 99-107 months Cycle 5 BMI  In the fourth phase, bivariate analysis, using ANOVA, was performed to identify the associations between early childhood temperament and parenting practices and membership in the different BMI trajectory groups. The final phase, using discriminant analysis, identified early childhood temperament and parenting practices that distinguished membership in the different BMI trajectory groups (i.e., to identify which parenting practices and children’s characteristics predicted membership in the identified BMI trajectory groups). The objective of the analyses was to test the fit between parenting practices and children’s temperament and their interactive contribution to obesity. This phase was based on Tabachnick and Fidell’s (2007) description of the discriminant function model. Growth curve models This longitudinal study involved the identification of body mass index (BMI) trajectories of children that reflected their developmental change over time. The variables parenting practices and temperament were measured in conjunction with BMI. In studies such as this, where there is a need to account for the variability related to individual differences (or different data collection periods), and the intent is to study predictors of those individual differences and changes at the group level, it is best to use growth curve methods such as latent growth curve models (LGC) (Stoolmiller, 1995), group-based modeling approaches (Nagin, 1999, 2002), or hierarchical linear models (HLM) (Heo et al., 2003). According to Nagin (1999), group-based models assume that the population is constructed of a combination of distinct groups identified by their developmental trajectories  69 (based on the assumption of heterogeneous rather than homogenous population trajectories (Nagin, 2002)). The parameters of the group-based model are estimated by maximum likelihood estimation. LGC and HLM, on the other hand, assume a continuous distribution of trajectories within the population and do not identify distinctive clusters of trajectories. These latter two models (LGC and HLM) are best suited in studies where there is a desire to describe the continuous variability in the patterns of growth throughout the population. The group-based trajectory model uses a multinomial modeling strategy rather than the multivariate continuous distribution strategy used by both the HLM and LGC models (Lacourse, Nagin, Tremblay, Vitaro, & Claes, 2003). The model: A finite mixture model The group-based model that was applied had the potential to identify distinctive groups of developmental trajectories of BMI, over 6 years, in children that were 24 to 35 months of age at baseline. The analyses were planned to: (a) answer the question, “Do certain types of children have more distinctive BMI trajectories?” (b) estimate the proportion of the population of each such group with a particular BMI trajectory; and (c) use the BMI trajectory groups to produce profiles of group members. The BMI trajectories were established by fitting the “semi-parametric mixture model” to the NLSCY data. The model is a mixture of probability distributions that are properly defined to describe the data under investigation. This group-based method was employed to describe the course of BMI change for both boys and girls. The mixture model analysis - the estimation of the trajectories - was conducted by applying PROC TRAJ, in SAS® 9.1, which was developed by Jones, Nagin, and Roeder (2001); it is a procedure that is based on a semi- parametric, group-based modeling strategy.  70 Nagin and Tremblay (2004) described the group-based model as an application of “finite mixture modeling,” which is an extension of the maximum likelihood models. The use of the words finite mixture refers to the assumption that the population “comprises a mixture of a finite number of unobserved groups” (Nagin, 2005, p. 62). The group-based model assumes that the population is composed of a combination of groups with different developmental trajectories. Membership at the level of the individual is not observed and the estimation, which is based on mixture modeling of each group’s trajectory, is accomplished by identifying the shape of the trajectory for each group and the proportion of the population making up each of those groups (Nagin & Tremblay, 2001). The semi-parametric, group- based method utilizes a multi-nomial function to model the relationship between the variables (Nagin, 1999). The type of data to be analyzed grants different forms to the group-based trajectory models: “the technical specifics of the statistical model used to identify and estimate the trajectory groups depend on the form of the response variable under investigation” (Nagin & Tremblay, 2001, p. 21). For this study, the trajectory groups were identified and estimated by using the censored normal model (CNORM); CNORM is typically used to model the conditional distribution of a censored variable where there is a cluster of data at the maximum or minimum values (Jones et al., 2001), or for data that are measured on a continuous scale without censoring (e.g., BMI) (Nagin, 2005). The relationship between time and the behavior of interest (in this instance BMI was the variable of interest, which is not a behavior, per se) is established by way of a latent variable, y*j it, which is assumed to be a measure of the “individual i’s potential for engaging in the behavior of interest at time t” (Nagin, 2005, p. 28). The term latent in this context implies that the variable is not ‘fully  71 observed’ (Nagin, Tremblay, & Vitaro, 2003). Up to a fourth order polynomial relationship can be assumed between the latent variable and time, which in this study was measured as age in months (Jones & Nagin, 2005). The relationship between the behavior or attribute of interest and age is described in the following form: y*j it = Β0j + Β1j Age it +  Β2j Age it2 +  Β3 j Age it3 +  Β4 j Age it4 +  εit , where the latent variable y*j it characterizes the level of the behavior or attribute for individual i at time t given membership in group j; εit is an error term that is assumed to be normally distributed with a zero mean and a constant standard deviation, σ; Age it, Age it2, Age it3, and Age it4  are the ages of the child i at time t, the squared age of the child i at time t, the cubed age of the child i at time t, and age to the fourth power of child i at time t, respectively; and Β0j , Β1j , Β2j, Β3 j, and  Β4 j are the parameters that determine the shape of the trajectory for group j, also described as the model’s coefficients. Given that separate sets of parameters are estimated for each group j, the coefficients are free to vary across groups (Nagin, 2005). Accordingly, this freedom allows for the detection of population heterogeneity, or distinctly different developmental paths, at the level of the attribute or behavior under study at a given age, and its development over time (Nagin, Pagani, Tremblay, & Vitaro, 2003). The identification of the distinct groups of individual developmental trajectories involved the initial steps of: (a) model selection through establishing the optimal number of groups and trajectory shapes that best fit the data (polynomial function) and (b) estimating the proportion of individuals in each of the groups. Once these two steps were completed, the analysis progressed to (c) the identification of the probability of each individual belonging to each of the trajectories (i.e., posterior group-membership probability). Figure 1 provides an overview of the group-based model.  72 Figure 1. Overview of the model (adapted from Nagin, 1999)                    Model selection Model selection involved two components, one was the selection of the most empirically appropriate number of groups J to construct the mixture, and the second was determining the most appropriate order of the polynomial applied to model the trajectories. Optimal number of groups  For the purpose of selecting the optimal number of groups J, a succession of models was run. To evaluate alterations in model fit, the models were compared based on the change in the log-likelihood and the Bayesian Information Criterion (BIC) (Jones et al., 2001). The BIC is considered to be a conservative criterion for model selection because it tends to support a smaller number of groups than the true number. Further, the performance of the BIC improves with sample size (Nagin, 2005). It is recommended that when prior  DATA Optimal number of groups and trajectory shapes Proportion of population in each group Probability that individual i belongs to trajectory group j  73 information of the correct model is limited (i.e., empirical and theoretical), model selection is based on the maximum BIC (i.e., the least negative BIC value) (Nagin, 1999), which tends to support more parsimonious models (Jones et al., 2001). For a particular model, the BIC is calculated as: BIC = log(L) - 0.5klog(N), where N is the sample size, L is the value of the model’s maximized likelihood, and k is the number of parameters in the model (Nagin, 1999), where “the number of the parameters is determined by the order of the polynomial used to model each trajectory and the number of groups” (Nagin, 2005, p. 64). The value of the BIC is always negative; therefore, the maximum BIC value will be the least negative. The closer the BIC is to zero, the stronger is the fit of the model to the data. Conversely, more negative BIC values reflect weaker model fit and negative changes in the BIC indicate a decline in fit (Nagin, 1999).  To evaluate the magnitude of the change in the BIC of competing models, one can refer to a metric developed by Kass and Wasserman (1995) and Shwarz (1978) and recommended for use by Nagin (1999). This metric, the Bayes factor (Bij), formulated from Bayesian statistics, measures the “posterior odds of i being the correct model given the data” (Nagin, 2005, p. 68). Use of this statistic allows for appraisal of the difference in the BIC scores of two models to points of reference known as Jeffreys’ scale of evidence for Bayes factors. The complexity of calculating the Bayes factor has led to an approximation of the Bayes factor (Nagin, 1999, 2005). Through this metric (i.e., BIC-based probability approximation), one can compare competing models based on the calculation of the probability statistic (Pj): Pj = e (BICj - BICmax)/∑j e(BICj - BICmax),  74 where BICmax is the maximum BIC score of the model under consideration. The model with the largest (closest to 1) BIC-based probability approximation is considered to be the most accurate model. A further criterion based on the change in the model BIC is the use of the BIC as an approximation of the log of the Bayes factor. This estimation is appropriate for testing the number of components in a mixture (Jones et al., 2001). The Bayes factor (Β10) provides the “posterior odds that the alternative hypothesis is correct when the prior probability that the alternative hypothesis is correct equals one-half” (Jones et al., 2001, p. 390). The BIC log Bayes factor approximation is as follows: 2loge(Β10) ≈ 2(∆BIC), the ∆BIC is the BIC of the alternative more complex model minus the BIC of the simpler model. Jones et al. (2001) described the log form of the Bayes factor as the “degree of evidence favoring the alternative model” (p. 390). The interpretation of 2loge(Β10), adapted from Jones et al. (2001), is summarized in Table 8. Table 8. Interpretation of 2loge(B10)  2loge(Β10) (Β10) Evidence against the null hypothesis 0 to 2 2 to 6 6 to 10 > 10 1 to 3 2 to 20 20 to 150 > 150 Not worth mentioning Positive Strong Very strong   75 Specifying the polynomial function  An objective of this modeling exercise was to estimate a set of parameters that could describe the shapes of the trajectories; which are described by a polynomial function of age or time with each polynomial function corresponding to a trajectory group (i.e., the specification of the order of the polynomial used to signify the shape of each group’s trajectory) (Nagin, 2005). Through this process, the model selected, based on the changes in the BIC, is refined by altering the order of the trajectories. There are many possible combinations of different ordered trajectories for a model (depending on the number of groups). According to Jones (2005), when theory provides little information about the shape of the trajectory, a strategy to determine the most appropriate order of the polynomial is to start with a third-order polynomial (i.e., cubic trajectories). The order for each group is then decreased until the parameter estimates are significant for each group. The best order for each of the trajectories is decided upon by the lowest significance statistic. Probability of group membership  The probability of group membership is defined as the “proportion of the population following each trajectory group j” (Nagin, 1999, p. 151). The probability of group membership, π j, is not calculated directly but rather is estimated by a multinomial logit function. The estimation guarantees that each group probability will be situated between zero and one and further ensures that across the J groups, the values of πj sum to one. This therefore permits “group membership probabilities to depend on characteristics of the individuals” (Nagin, 2005, p. 41). The estimations of group membership probabilities were obtained through the function:  76 ∑= jj jj ee 1 / θθπ , where θj is normalized to zero. Calculation of posterior group membership probabilities  The strength of mixture group modeling is the possibility of calculating the probability of membership of each individual into his or her trajectory group (Nagin, 1999). The probability of an individual i belonging to group j is referred to as the posterior probability of group membership and these probabilities are computed using the model’s estimated coefficients (Nagin, 2005). A posterior probability accounts for the observed outcomes of individual i over the t assessment periods (Nagin, 2005). Posterior probabilities are based on the maximum likelihood estimates of the trajectory parameters and are based on the individual longitudinal pattern of the behavior or attribute, Yi (Nagin, 1999). The notion of applying maximum likelihood is to select parameter estimates for which the probability of observing the actual data Y is maximized. In addition to establishing profiles of the trajectory group members, posterior probabilities can be used to evaluate the quality of the model’s fit to the data (Nagin, 2005). The posterior probability of individual i membership belonging to group j is represented by ( )i/YjP∧ , where Yi is a vector comprising i’s measured behavior or attribute in each assessment period t, yit (Nagin, 2005). The equation for calculating the posterior group membership probabilities is presented as follows: ( ) =∧ i/YjP ∧∧∧∧ ∑ ji j jYPjYP ji ππ )/(/])/([ “Where ( )i/YjP∧ is the estimated probability of observing i’s actual behavioral trajectory, Yi, given membership in j, and πˆj is the estimated proportion of the population in group j”  77 (Nagin, 1999, p. 149).51  Group assignment based on the posterior probability assignment rule is referred to as the average posterior probability of assignment (i.e., the AvePP). The closer the average posterior probability of assignment is to 1, the closer is the correspondence of the model with the data. The minimum rule of thumb for an acceptable average posterior probability (AvePP) is at least .70 for all groups (Nagin, 2005). Model adequacy In the following section, I describe Nagin’s (2005) recommendations for judging the adequacy of the models (i.e., evaluating the models’ ability to identify distinct groups of BMI trajectories for boys and girls). Model adequacy or fit is evaluated through the diagnostics, including the average assignment probability for the data, the odds of correct classification, the confidence intervals (CI) for the group membership probabilities, and the estimated group probabilities. Diagnostic 1: The average posterior probability of assignment  The first diagnostic relies on the average posterior probability (AvePP) for each trajectory group, which provides information about the correctness of the group membership classifications based on the maximum posterior probability assignment rule (Nagin, 2005). As described earlier, the closer the assignment probability is to 1, the closer is the model to the data. Further, the minimum rule of thumb for an acceptable (AvePP) is at least .70 for all groups (Nagin, 2005).  51 The posterior probability calculation is more likely to give higher probabilities to larger groups (Nagin, 2005).  78 Diagnostic 2: The odds of correct classification  The ‘odds of correct classification’ diagnostic also is based on the maximum posterior probability assignment rule (i.e., AvePP for group j and πJ). The ‘odds of correct classification’ statistic is computed as: ( )[ ] ⎥⎥⎦ ⎤ ⎢⎢⎣ ⎡ ∧− ∧ −= jjAvePP jAvePP jOCC j ππ 11  where πˆj is the probability of group membership or the proportion of the population following each trajectory j, [AvePPj / (1 - AvePPj )] is the odds of correct classification into group j, which is based on the rule of maximum probability classification, and ⎥⎦ ⎤⎢⎣ ⎡ − ∧∧ jj ππ 1 is the odds of correct classification based on random assignment. According to Nagin (2005), a value of OCCj that is larger than 5.0 for all groups points to high assignment accuracy of the model. However, Nagin (2005) added that if the maximum probability assignment rule (AvePP j) has no predictive capability beyond arbitrary chance, then the OCCj is equal to 1. Diagnostic 3: Confidence intervals for the predicted point estimates of the trajectories A third model diagnostic procedure that was presented by Nagin (2005) and Jones and Nagin (2005) is the confidence interval of the estimates of the groups’ predicted trajectory by age. Confidence intervals are calculated to assess the impact of sampling error on the accuracy of the statistical estimates (Nagin, 2005). A confidence interval that is narrow indicates that the point estimates are correctly estimated (i.e., the trajectories are distinct) while a wide interval means that the point estimates are questionable (i.e., the  79 trajectories may overlap). Diagnostic 4: Estimated group probabilities versus the proportion of the sample assigned to the group The fourth diagnostic tool relies on the proportion of the sample assigned to group j (Pj) on the basis of the maximum posterior assignment rule, where Pj represents the proportion of the sample assigned to group j. This proportion equals N j / N, “where N j is the number assigned to group j and N is the total sample size” (Nagin, 2005, p. 89). If the group’s AvePP of assignment is 1, indicating perfect group assignment, then π j and Pj are identical. As incorrect assignment increases, the association between j ∧π  and P j deteriorates. Discriminant analysis Discriminant analysis is used to establish variables that discriminate between two or more variables. The application of this analysis assists in predicting group membership from a number of predictors and to identify whether group membership is related to a statistically significant difference in a dependent variable score (Tabachnick & Fidell, 2007). The same research question could be explored through logistic regression analysis although a decision on which method to apply depends on the distribution and relationship among the predictors and the distribution of the dependent variable. Once the assumptions of normality, linearity, and homogeneity are met, the discriminant function is superior to logistic regression. This method has more statistical power compared with logistic regression and therefore reduces the probability of Type II error. The advantages in using this analytical method are that it: (a) permits the interpretation of the pattern of differences among the predictors as a whole (i.e., the theoretical notion of goodness-of-fit between the variables of interest could be examined)  80 when a significant association is detected between the predictors and the dependent variables (Tabachnick & Fidell, 2007) and (b) provides more accurate classification and hypothesis testing (Grimm & Yarnold, 1995). Discriminant analysis proceeds in two steps, first by using Wilks’ Lambda (i.e., F test) to assess the overall significance of the discriminant model and second by assessing the individual predictors to determine whether they significantly differ, on average, by group. Variable selection and model building Variables considered in these analyses were based: (a) on the hypothesized explanatory variables (i.e., temperament and parenting practices) and (b) relevant confounding variables including the age, income and level of education of the PMK. For the purpose of this study, direct discriminant analysis was applied. This is the preferred method when there is no specific order or importance for the predictors and consequently the predictors can be concurrently entered into the equation (Tabachnick & Fidell, 2007). The analysis was conducted twice; first a direct main effects model was estimated and second the direct main effects model was adjusted by introducing the confounding variables, selected family characteristics that have been identified to influence parenting practices and children’s weight status. Missing data Although the SAS procedure Proc Traj is specified to handle missing data, only cases with at least three measures of BMI, over the four time points of measurement, were included in the analysis. To account for the complexity of the survey’s design, including both clustered and stratified sampling, standardized sample weights were used in all analyses. The  81 standardized weights were calculated by dividing the longitudinal weight provided by Statistics Canada over the average of that weight for the boys and girls, separately.       82 Chapter 4: Results This chapter is divided into seven major components: (a) a general overview of the merging and inclusion procedures for the children from Cycles 2-5 of the NLSCY and the ensuing study cohort, (b) the results of the attrition analysis of the included and excluded children, (c) a general description of the sample: geographic distribution, family demographics, and the children’s health status, (d) the results of the univariate analyses of the children’s BMI, their reported temperament, and the PMKs’ reported parenting practices, (e) the results of the Proc Traj modeling and a description of the resultant BMI trajectories, (f) the associations between the BMI trajectory groups and the primary research variables identified through ANOVA, and (g) the results of the goodness-of-fit modeling conducted through direct discriminant analysis. The participants (study cohort) The participants for this study were drawn from the NLSCY; these children had repeated measures in the Cycles 2 to 5 datasets. Children between the ages of 24 to 35 months in Cycle 2 of the NLSCY were selected as the baseline. In the subsequent cycles, only children with data in the second cycle (i.e., baseline) were included in the longitudinal file. The second inclusion criterion depended on the derived outcome variable BMI. The calculation of BMI was contingent on two variables being present, the weight of the child and the height of the child. As a result, children missing either one of these variables did not have a BMI variable. Of the 1,890 cases in Cycles 2, 3, 4, and 5, only 1,623, 1,340, 960, and 1,030 cases, respectively had a calculated BMI.  83 When analyzing childhood developmental characteristics, such height, weight, or BMI, the identification of outlier observations is crucial to the quality of the analysis. Outliers for height and weight data are described as “biologically implausible values” (BIVs) (CDC, 2005). It is assumed that the outliers in these cases are due not to actual excessive growth, but are the result of inaccurate measurements, errors in data entry, or in this case, in the reporting of height or weight by the PMK. For a full description of World Health Organization and Centers for Disease Control definitions of outlier cutoffs, the reader is referred to WHO (1995). The outlier analysis was undertaken with a SAS® program that identified outlier observations (i.e., BIVs). This SAS program identified the extremely low and high BMI values according to the WHO fixed exclusion ranges (the program was downloaded from: http://cdc.gov/nccdphp/dnpa/growthcharts/sas.htm). Following on these recommendations, the numbers of flagged BMI values per cycle are described in Table 9. Table 9. Cases with outlier BMI values per survey cycle  Cycle N with calculated BMI Acceptable range BMI (N) Extremely low BMI (N) Extremely high BMI (N) Missing 2 1623 1385 12 226 267 3 1340 1234 31 75 550 4 960 931 13 16 930 5 1030 1011 6 13 86   84 From Table 9, 392 cases had biologically implausible values. Of note, the program identified higher numbers of extremely high BMI values rather than extremely low BMIs. To obtain reliable parameter estimates of the trajectories, it is recommended that each case has a minimum of three time-points of measurement (Nagin, 2005). Accordingly, only cases with a minimum of 3 measures of acceptable range BMI were included in the trajectory analysis. This resulted in a total sample of 972 cases included in the analysis. Of these cases, there were 490 girls and 482 boys. The break down in the sample size from the point of merging with 1,890 cases to the final sample size of 972 is presented in Table 10. Table 10. Counts of cases with acceptable BMI values  N Total N Girls N Boys Number of Cases with Acceptable Range BMI (N = 1890)  110 370 438 573 399 59 171 222 282 208 51 199 216 291 191 No acceptable BMI values One acceptable BMI value Two acceptable BMI values Three acceptable BMI values Four acceptable BMIs values  Attrition analysis  The procedures applied above in determining the final study sample resulted in a sample (n = 972) that was considerably smaller than the total cohort of children between the ages of 2-3 years (n = 1890). This introduced a concern about whether the participants who were excluded were at all different from the 972 children included in the study. To address  85 the possibility of selection bias, an attrition analysis was conducted of the major variables (on several family and child characteristics); the children who were excluded from the study were compared with the included participants. Specifically, the analysis was conducted using chi-square analysis for the categorical variables: BMI classification (prevalence of overweight and obesity), age of the PMK at the time of the interview, household income, and the educational attainment of the PMK. Further analysis, using ANVOA, was conducted for the continuous variables: consistency in discipline, positive interactions, ineffective parenting, and children’s reported temperament.  Results from the attrition analysis revealed that the children who were excluded differed significantly from those that were included. When compared with the included children, the children who were excluded were more likely to be from low income homes, and had PMKs with lower educational attainments. Their PMKs were more likely to be in the 15-29 year age range and were also significantly more inconsistent in their disciplinary practices. Across the four cycles, the children who were excluded had higher rates of obesity (classification based on Cole’s et al. cutoffs) when compared with the BMI classification of the included children. The excluded children did not differ from the included participants in their reported temperament or in their PMKs’ positive or ineffective parenting practices. Sample description In the following analyses, the boys and girls data were weighted with a standardized weight. The standardized sample weight was calculated by dividing the longitudinal weight provided by the NLSCY by the mean of that longitudinal weight for the 482 boys and 490  86 girls, separately.52 The following section provides the geographic distribution of the participants, a description of the demographic characteristics of the PMKs and their spouses, and the children’s health status. The descriptive analyses were conducted with SPSS 14. Geographical characteristics of the participants Table 11 shows that the majority of the sample resided in central Canada (60.8% of the boys and 62.9% of the girls), whereas only 7.3 % of the boys and 8.1% of the girls resided in Atlantic Canada. Almost one half of the children lived in urban areas that were populated with 500,000 or more people (43.4% of the boys and 46.0% of the girls) with only 14.1% of the boys and 12.2% of the girls residing in rural areas. Table 11. Geographical characteristics  Variable Boys (n = 482) Girls (n = 490)  Frequency (%) Frequency (%) Region of residence Atlantic Canada Central Canada The Prairies Pacific Canada 35 (7.3%) 293 (60.8%) 96 (19.9%) 58 (12.0%) 39 (8.1%) 308 (62.9%) 83 (16.9%) 59 (12.1%) Area size of residence Rural area Urban, population < 30,000 68 (14.1%) 70 (14.5%) 60 (12.2%) 66 (13.5%)  52 Standardized sample weight = Longitudinal weight (calculated by Statistics Canada divided by the mean of the longitudinal weight.  87 Variable Boys (n = 482) Girls (n = 490)  Frequency (%) Frequency (%) Urban, population 30,000 to 99,999 Urban, population 100,000 to 499,999 Urban, population 500,000+ 46 (9.6%) 83 (17.2%) 209 (43.4%) 36 (7.3%) 99 (20.3) 225 (46.0%)   Demographic characteristics of the responding parents (surrogates) and their spouses Table 12 summarizes the descriptive statistics of the PMKs and their spouses, at baseline. For the total sample, the majority of the PMKs ranged in age from 30 to 34 years, and only 5.5% and 8.7% of the boys’ and girls’ PMKs were between 15 and 24 years of age, respectively. More of the boys’ PMKs reported their age to be over 40 years; only 3.9% of the girls’ PMKs were over 40 years. The age distribution of the PMKs’ spouses was similar to that of the PMKs’. The majority of the spouses ranged in age from 30 to 35 years. Over 69% of the boys’ and about 71% of the girls’ birth mothers were between 25 and 34 years of age when the children were born. Approximately 50% of the boys’ PMKs had a college or university education and only 5.8% held less than secondary school education. Slightly less than one half (46%) of the girls’ PMKs had completed a college or university degree and 9.5% had less than secondary school education. The majority of the children in the study were in households with total income equaling or exceeding $40,000 per year (67.8% of the boys and 60.9% of the girls). About 5% of the children lived in households with less than $15,000 per annum. The  88 majority of PMKs rated their general health to be very good or excellent; about 4% rated their health as fair to poor. Table 12. Parents’ (surrogates’) demographics   Boys (n = 482) Girls (n = 490) Variable Frequency (%) Frequency (%)  Age group of PMK 15-24 years 25-29 years 30-34 years 35-39 years 40+ years  27 (5.5%) 99 (20.6%) 219 (45.5%0 104 (21.5%) 33 (6.9%) 43 (8.7%) 118 (24.1%) 193 (39.4%) 117 (23.8%) 19 (3.9%) Age group of spouse 15-24 years 25-29 years 30-34 years 35-39 years 40+ years  8 (1.6%) 55 (11.4%) 189 (39.2%) 143 (29.6%) 50 (10.3%) 14 (2.8%) 73 (14.8%) 166 (33.9%) 141 (28.7%) 45 (9.3%) Age group of mother at birth 15-24 years 25-29 years 30-34 years  62 (13.0%) 167 (34.7%) 167 (34.7%) 78 (16.0%) 174 (35.6%) 173 (35.3%)  89  Boys (n = 482) Girls (n = 490) Variable Frequency (%) Frequency (%)  35-40+ years 80 (16.6%) 58 (11.8%) PMK’s highest level of schooling Less than secondary school graduation Secondary school graduation Beyond high school College or university degree (including trade)  28 (5.8%) 85 (17.6%) 130 (27.0%) 239 (49.5%) 47 (9.5%) 69 (14.0%) 149 (30.4%) 226 (46.0%) Spouse’s highest level of schooling Less than secondary school graduation Secondary school graduation Beyond high school College or university degree (including trade)  51 (10.5%) 65 (13.4%) 93 (19.4%) 226 (46.9%) 52 (10.6%) 79 16.1%) 113 (23.0%) 182 (37.2%) Household income Less than $10,000 $10,000 to $14,999 $15,000 to $19,999 $20,000 to $29,999 $30,000 to $39,999 $40,000 or more  5 (.8%) 27 (5.4%) 18 (3.8%) 41 (8.6%) 65 (13.6%) 237 (67.8%) 5 (1.1%) 17 (3.5%) 23 (4.7%) 56 (11.5%) 90 (18.3%) 298 (60.9%) PMK’s general health status Excellent  211 (43.8%) 195 (39.9%)  90  Boys (n = 482) Girls (n = 490) Variable Frequency (%) Frequency (%)  Very good Good Fair to poor 163 (33.9%) 83 (17.2%) 25 (4.5%) 188 (38.3%) 90 (18.4%) 16 (3.4%)  The children’s health status Table 13 presents frequencies and percentages related to the health of the children. The majority of the boys and girls were reported to experience very good to excellent health (approximately 89% of the boys and 93% of the girls). Conversely, only 2% of the boys and 1% of the girls were reported to experience fair to poor health. When asked about the activity level of the child compared to others, the PMKs reported that 98% of the boys and about 99% of the girls were equally or more active than others. Around 21.0% of the girls were reported to be ‘much more active than others’, only 13.7% of boys were described as such. The data showed generally healthier girls than boys. Over 90% of the boys and girls were reported to never have had asthma. More boys (9.5%) than girls (6.7%), however, had experienced asthma. Furthermore, around 11% of the boys and 4% of the girls were reported to have had allergies or to have used inhalants (6.4% boys and 3.8% girls). Moreover, 19% of the boys and 9.4% of the girls had a chronic condition at Cycle 2. About 6% of the girls and 8% of the boys were reported to use prescription medication on a regular basis.  91 Table 13. Children’s health status  Variable Boys (n = 482) Girls (n = 490)  Frequency (%) Frequency (%) Child’s health status Excellent Very good Good Fair or poor 293 (60.9%) 135 (28.0%) 43 (9.0%) 10 (2.0%) 310 (63.3%) 143 (29.2%) 32 (6.5%) 5 (1.0%) Activity of child compared to others Much more Moderately more Equally Moderately less 66 (13.7%) 139 (28.7%) 269 (55.7%) 9 (1.8%) 103 (21.0%) 105 (21.4%) 276 (56.4%) 5 (1.1%) Ever had asthma Yes No 46 (9.5%) 436 (90.5%) 33 (6.7%) 457 (93.3%) Allergies Yes No 52 (10.8%) 430 (89.2%) 19 (3.9%) 471 (96.1%) Used Asthma Inhalers* Yes No 31 (6.4%) 451 (93.6%) 19 (3.8%) 471 (96.2%) Chronic condition  92 Variable Boys (n = 482) Girls (n = 490)  Frequency (%) Frequency (%) Yes No 91 (18.9%) 391 (81.1%) 46 (9.4%) 444 (90.6%) Used prescription drug regularly Yes No 38 (8.0%) 444 (92.0%) 31 (6.2%) 459 (93.8%)   * Does the child take Ventolin, inhalers or puffers for asthma on a regular basis?  Univariate analysis The following sections present the results of the univariate analysis for the primary variables BMI, children’s reported temperament and the three parenting practices variables: consistency in discipline, positive interactions, and ineffective parenting. BMI The BMI values were calculated for Cycles 2 and 3 by dividing the child’s reported weight over the child’s reported height squared. The BMIs for Cycles 4 and 5 were provided in the dataset by Statistics Canada. Table 14 shows the girls’ and boys’ BMI descriptive statistics for Cycles 2 through 5. The boys had the fewest cases with missing values (n = 37) in Cycle 2 and the greatest number in Cycle 5 (n = 121). The girls, on the other hand, had the most missing data in Cycle 4 and the least in Cycle 2.      93 Table 14. Descriptive statistics for the girls’ and boys’ BMI   Cycle 2 (n = 440 Girls) (n = 445 Boys) Cycle 3 (n = 448 Girls) (n = 435 Boys) Cycle 4 (n = 377 Girls) (n = 396 Boys) Cycle 5 (n = 403 Girls) (n = 361 Boys) Mean Girls Boys  17.60 17.56  16.94 16.84  17.18 17.06  17.41 17.46 Standard error of mean Girls Boys   .13 .12   .13 .12   .18 .17   .16 .18 Standard deviation Girls Boys  2.75 2.47  2.75 2.46  3.56 3.29  3.24 3.37 Skewness Girls Boys  .34 .46  .68 .48  .84 .95  .94 1.11 Kurtosis Girls Boys  -.39 -.16  .18 -.20  .16 .25  1.15 1.47 Missing  94  Cycle 2 (n = 440 Girls) (n = 445 Boys) Cycle 3 (n = 448 Girls) (n = 435 Boys) Cycle 4 (n = 377 Girls) (n = 396 Boys) Cycle 5 (n = 403 Girls) (n = 361 Boys) Girls Boys 50 37 42 47 113 86 87 121   The means and standard deviations for the boys’ and girls’ BMIs were relatively comparable. The average BMI was highest during Cycle 2 for both the boys and the girls (17.6 and 17.6, respectively). Conversely, the average BMI was lowest for both the girls and the boys during Cycle 3 (16.8 for boys and 16.9 for girls). The girls seemed to have greater variability in their BMI compared with the boys. The variability was noted to be relatively higher in Cycles 4 and 5 compared with Cycles 2 and 3 for both the boys and the girls.  The skew and kurtosis statistics for Cycles 2 through 4 of the boys’ data were less than 1 indicating that the BMI variables for the three cycles were normally distributed. The skew and kurtosis statistics for Cycle 5 was greater than 1 but less than 2, which is the maximum acceptable value to support a claim of normality (Miles & Shevlin, 2004). The distributions of Cycles 2, 3 and 4 contained few outliers, although Cycle 5 contained some extreme outliers. The skew and kurtosis values for the girls’ data for Cycle 2 were all within acceptable range and supported a normality claim. The distributions for the four cycles were slightly skewed; Cycle 4 was the most skewed while Cycle 3 was the least. Cycles 3 and 4 had some  95 outliers although not as extreme as the outliers observed in Cycle 5. Table 15 provides the boys’ and girls’ prevalence rates of overweight and obesity, using Cole’s et al. cutoff points, for Cycles 2 through 5. The percentage of valid cases was greatest in Cycles 2 and 3 for both the girls (95.4% and 92.2%) and the boys (95.7% and 93.9%). The greatest prevalence rate of overweight was observed in the girls during Cycle 3. The table also shows that obesity was more common in girls than boys in Cycles 2 through 4. The rate of overweight among boys remained fairly constant over time and was lower than the rate of obesity in the first three cycles. The lowest rate of obesity occurred in boys during the fifth cycle (9.3%).  Table 15. Prevalence rates of overweight and obesity based on Cole's cutoffs  Cycle Boys (N = 482) Girls (N = 490)  N (%) Normal weight n (%) Overweight n (%) Obesity n (%) n (%) Normal weight n (%) Overweight n (%) Obesity n (%) 2   3   4 461 (95.7%)  452 (93.9%)  398 306 (63.6%)  308 (63.9%)  265 60 (12.5%)  63 (13.2%)  61 94 (19.6%)  81 (16.9%)  72 468 (95.4%)  452 (92.2%)  379 261 (53.3%)  265 (54.1%)  234 84 (17.1%)  85 (17.3%)  62 123 (25.1%)  102 (20.8%)  84  96 Cycle Boys (N = 482) Girls (N = 490)  N (%) Normal weight n (%) Overweight n (%) Obesity n (%) n (%) Normal weight n (%) Overweight n (%) Obesity n (%)   5 (82.5%)  362 (75.2%) (54.9%)  257 (53.3%) (12.7%)  61 (12.6%) (14.9%)  45 (9.3%) (77.4%)  410 (83.7%) (47.8%)  296 (60.5%) (12.6%)  69 (14.1%) (17.0%)  44 (9.0%)   Temperament The results of the factor analysis conducted by Statistic Canada yielded a factor that was labeled the ‘fussy/difficult’ temperament dimension. To confirm that the same results would apply to this cohort, a principal component analysis was conducted with the 10 items for the 1,890 cases in Cycle 2 using the default eigenvalue of one. This resulted in a 2-factor solution. Factor one included the 8 items: “easy/difficult to calm child,” “at times child gets fussy,” “frequency of crying,” “how easily the child gets upset,” “how vigorous the child cries,” “mood changeability,” “amount of attention required,” and “the overall degree of difficulty.” All the loadings were above .3. The items, “general mood of the child” and “times the child makes happy sounds and laughs” loaded on the second factor with loadings of .84 and .86, respectively. Consequently, a second factor analysis was conducted with the  97 item “child smiles, makes happy sounds, laugh” removed. The results showed one clearly defined factor with all items loadings above .3. This one item was thus removed from the factor; it was likely to have functioned poorly because children who are reported to be difficult are also known to experience periods of vigorous laughter and happiness (Thomas & Chess, 1977). Cronbach’s alpha for the 9-item scale was .81, reflecting a satisfactory degree of internal consistency.  To construct the variable, fussy/difficult temperament, all 9 items (each item rated from 1 to 7) were summed to form a temperament variable with values that ranged from 9 to 63. Scores of 9 indicated that the child was judged to be low in fussiness and difficulty while scores nearing 63 indicated that the child was extremely fussy and difficult. The derived variable is described in Table 16. Table 16. Temperament: Descriptive statistics   Mean Std. error of mean Standard deviation Skewness Kurtosis Boys (n = 480) 27.2 .41 8.9 .24 -.20 Girls (n = 490) 27.2 .35 7.7 .10 .44  The average difficult/fussy temperament scores for the boys and the girls were similar although the boy had more variability (SD = 8.9 vs. girls’ SD = 7.7). The skewness and kurtosis statistics, and visual inspection, indicated that the variables were normally distributed. There were some outliers although no extreme outliers.  98 Parenting Table 17 presents the descriptive statistics for the positive interaction, ineffective parenting and consistency parenting practices variables for both the boys and the girls. Table 17. Parenting: Descriptive statistics  Variable Mean Std. error of mean St. dev.  Skewness Kurtosis Positive interaction Boys (n = 478) Girls (n = 486)  17.0 16.8  .10 .11  2.2 2.4  -.60 -.61  .14 -.18 Ineffective parenting Boys (n = 477) Girls (n = 484)  9.2 8.9  .16 .18  3.9 4.0  .31 .65  -.02 .29 Consistency Boys (n = 475) Girls (n = 482)  14.8 14.2  .16 .16  3.4 3.4  -.71 -.58  .18 .23  Table 17 illustrates that the means of the parenting variables for the boys were similar to those of the girls. The data revealed more variability in the parenting variable scores for the girls compared with the boys. The skewness and kurtosis statistics indicated mildly skewed distributions for all three parenting variables, although they were within acceptable range; all skewness and kurtosis values were below 2. By examining the box plots, outliers  99 were present in the left of the distributions for positive parenting and consistency meaning that there were few parents who reported parenting practices that lacked positive encounters or consistency. On the other hand, the ineffective parenting boxplot revealed outliers on the right side of the distribution; these were the PMKs who reported lack of ineffective parenting styles. BMI trajectory modeling results Data organization and analysis   The first objective of this study was to describe and analyze the patterns of development of BMI of 2-year old children over a 6-year span applying the mixture-group model for developmental trajectories developed by Nagin (2005). To achieve the objective of estimating developmental trajectories, several successive models were built using the procedure Proc Traj, in SAS® 9.1.  In the following paragraphs, I present the results of the analysis including: (a) model selection based on finding the optimal number of groups and trajectory shapes that best fit the data (polynomial function), (b) the estimated proportion of individuals in each of the groups, and (c) the probability of each individual belonging to each of the trajectories (i.e., the posterior group-membership probability). Model selection Model selection involved two components, one was the selection of the most empirically appropriate number of groups to construct the mixture, and the second was determining the most appropriate order of the polynomial applied to model the trajectories.  100 The optimal number of groups For the purpose of selecting the optimal numbers of groups, a succession of models were run for the 482 boys and the 490 girls, separately. To evaluate the alterations in model fit, the models were compared based on: (a) the change in the log-likelihood Bayesian Information Criterion (BIC), (b) the BIC-based probability approximation, and (c) the BIC log Bayes factor approximation (Jones et al., 2001; Nagin, 2005). Girls’ Trajectories Based on the actual change in the BIC, the data from the NLSCY supported an all cubic 4-group model for the girls’ BMI trajectories. Table 18 reports the BIC scores for the models with 1 to 5 groups. Table 18. Selecting the number of groups based on the BIC (girls)  Number of groups BIC (N = 490)* BIC (N = 1678)** Probability correct model 1 -4247.79626 -4250.87365 0.00 2 -4176.63404 -4182.78881 0.00 3 -4168.21127 -4177.44341 0.00 4 -4156.57121 -4168.88074 0.96 5 -4159.6607 -4175.05297 0.04 *This N represents the number of girls in the estimation sample with 3 or more BMI measures. **The larger N pertains to the total number of assessments used in the model estimation across person and time (Nagin, 2005).53  53 N is meant to measure the number of independent observations that make up the sample. Because the intra-individual observations are not totally independent, the N across the individual and time overstates the theoretically correct smaller N (Nagin, 2005). To assess the model fit of groups-based models, two BIC statistics are reported, one defines the actual  101 Based on the application of the largest BIC score criterion, the girls’ data supported a 4-group solution, with BIC = -4156.57. Specifically, the BIC increased steadily becoming less negative as the number of groups increased from 1 to 4 groups and subsequently declined for the fifth group (BIC = -4159.66 ). Further support for the 4-group model was determined by the BIC-based probability approximation of .96, which is almost a perfect probability. Boys’ Trajectories Considering the change in the BIC values for the boys’ models (see Table 19), a 3- group solution offered the best fit to the data, with a BIC = -3993.75 compared with a BIC = -3993.99 for the 4-group model. The narrow change in BIC between the 3- and 4-group models was further manifested in the closeness between the two model probabilities. The probability of the 3-group model being the correct model was 0.56, which is a close competitor to the probability of the 4-group model (0.44). Further examination of the significance of the probabilities of group membership revealed that 3 out of 4 trajectories membership were not significant in the 4-group model, thus I opted for the 3-group model.        number of individual subjects in the sample (i.e., the number of girls), and the second larger N reports the data points used in model estimation (i.e., the number of observations for all subjects).  102 Table 19. Selecting the number of groups based on the BIC (boys)  Number of groups BIC (N = 482)* BIC (N = 1637)**  Probability correct model 1 -4073.83565 -4076.89234 0.00 2 -4025.06068 -4031.17406 0.00 3 -3993.75318 -4002.92325 0.56 4 -3993.98712 -4006.21389 0.44 *This N represents the number of boys in the estimation sample with 3 or more BMI measures. **The larger N pertains to the total number of assessments used in the model estimation across person and time.   The evaluation of the optimum number of groups based on the interpretation of the 2loge(Β10) in Table 20 is summarized in Tables 21 for the girls and 22 for the boys. Table 20. Interpretation of the 2loge(B10)  2loge(Β10) (Β10) Evidence against the null 0 to 2 2 to 6 6 to 10 > 10 1 to 3 2 to 20 20 to 150 > 150 Not worth mentioning Positive Strong Very strong      103 Table 21. Tabulated BIC and 2loge(B10) values for the girls' models  Number of groups BIC Null Model 2loge(Β10) 1 2 3 4 5 -4247.79626 -4176.63404 -4168.21127 -4156.57121 -4159.6607  1 2 3 4  142.32 16.84 23.28 -6.30   Table 22. Tabulated BIC and 2loge(B10) for the boys' models  Number of groups BIC Null Model 2loge(Β10) 1 2 3 4 -4073.83565 -4025.06068 -3993.75318 -3993.98712  1 2 3  97.55 62.62 -0.47  According to Jones et al.’s (2001) recommendation, and based on the interpretation of the BIC log Bayes factor approximation, the girls’ 4-group model is the most parsimonious fit (model) to the data. The change in BIC for the 4-group model was 23.28 compared with a negative change of -6.29 for the 5-group model. Similarly, the change in the BIC for the boys’ data favored a 3-group model as the most parsimonious fit (change of 62.62 compared with a negative BIC).   104 Summary of the optimal number of groups Based on the BIC calculations, the difference in the population distribution of BMI trajectories was best characterized by a 4-group model for the girls’ data and a 3-group model for the boys’ data. This was further supported by a comparison between the competing models based on the BIC-based probability of model correctness. For the girls’ data, the 4- group model had the highest BIC value and the probability of it being the correct model was 0.96 while a 3-group model best fit the boys’ data. Specifying the polynomial function As previously indicated, an objective of this exercise was to estimate a set of parameters that could describe the shapes of the trajectories. The shape of the trajectories is described by a polynomial function of age or time and each polynomial function corresponding to a trajectory group (Nagin & Tremblay, 2004). Based on Jones’s (2001) recommendations, the strategy to determine the most appropriate order of the polynomial was to start with a third-order polynomial (i.e., all cubic trajectories). The order for each group was decreased until the parameter estimates were significant for each group. The order for each of the trajectories was decided upon by the lowest significance statistics.  Polynomial function: Girls The following Proc Traj output (see Figure 2) presents the parameter estimates for the girls’ 4-group model with all-cubic trajectories.       105 Figure 2. Proc Traj output: Girls' 4-group model with all-cubic trajectories     Maximum Likelihood Estimates Model: Censored Normal (CNORM)  Standard       T for H0: Group   Parameter    Estimate        Error     Parameter=0    Prob > |T|  1  Intercept    39.58057     11.59135           3.415       0.0007  Linear       -1.26128      0.59474          -2.121       0.0341  Quadratic     0.02290      0.00863           2.654       0.0080  Cubic        -0.00012      0.00004          -3.125       0.0018  2 Intercept    18.73090      2.24094           8.358       0.0000  Linear       -0.05741      0.11954          -0.480       0.6311  Quadratic    -0.00028      0.00200          -0.141       0.8877  Cubic         0.00001      0.00001           0.600       0.5484  3 Intercept   -11.82252     14.98871          -0.789       0.4304  Linear        1.95615      0.79220           2.469       0.0136  Quadratic    -0.03452      0.01234          -2.797       0.0052  Cubic         0.00018      0.00006           2.982       0.0029  4        Intercept    21.57685     11.38298           1.896       0.0582  Linear       -0.26540      0.60866          -0.436       0.6629  Quadratic     0.00488      0.00956           0.510       0.6099  Cubic        -0.00002      0.00005          -0.442       0.6584        The significance of the cubic order parameters for groups one and three of the 4- group model indicated that a cubic specification was required to describe the shape of the BMI trajectories. However, the non-significance of the cubic specifications for groups two and four implied some degree of over fitting and that a higher order was not required to describe the shape of these trajectories. The second and fourth groups could be adequately specified by a zero-order function. I assigned different combinations of alternative polynomial specifications to the 4-group, all-cubic model. Table 23 presents the BIC values for the different competing models. For the girls’ data, the BIC values for the different model specifications showed that the refinements made by altering the polynomial function did not contribute to better fit. Consequently, the 4-group, all cubic model was the best model for the  106 girls’ data. Table 23 presents the various combinations with their corresponding BIC values. Table 23. BIC calculations for different models with combinations of alternative specifications for the girls' data Model Polynomial order BIC calculations 1 2 3 4 5 3, 0, 3, 0 3, 0, 3, 2 3, 3, 3, 0 3, 3, 3, 2         3, 3, 3, 3 -4170.97822 -4173.83196 -4164.37937 -4164.68541 -4156.57121   Polynomial function: Boys The following output provides the results of a 3-group all-cubic trajectory model fitted to the 482 boys’ data.                       107 Figure 3. Proc Traj output: Boys’ 3-group model with all-cubic trajectories    Maximum Likelihood Estimates Model: Censored Normal (CNORM)                                     Standard       T for H0:  Group   Parameter    Estimate        Error     Parameter=0   Prob > |T|   1       Intercept    16.71069      2.08245           8.025       0.0000          Linear        0.07244      0.11226           0.645       0.5189          Quadratic    -0.00231      0.00182          -1.271       0.2040          Cubic         0.00002      0.00001           1.702       0.0890   2       Intercept    34.96859      5.61600           6.227       0.0000          Linear       -1.09148      0.35117          -3.108       0.0019          Quadratic     0.02098      0.00626           3.350       0.0008          Cubic        -0.00012      0.00003          -3.529       0.0004   3       Intercept    30.22534      4.58174           6.597       0.0000          Linear       -0.48395      0.26840          -1.803       0.0716          Quadratic     0.00395      0.00460           0.858       0.3912          Cubic         0.00000      0.00002           0.122       0.9033      From the results (Figure 3), the significance of the cubic order of the 3-group model indicated that a cubic specification for group two was necessary to describe the shape of the BMI trajectory. Nonetheless, the non-significance of the parameter estimates for the remaining groups indicated a need to modify or refine the orders. Particularly, group one may have been better specified with a zero order and group three with a linear order as opposed to cubic orders. Consequently, I estimated the 3-group model with different ordered trajectories. Table 24 presents the various combinations with their corresponding BIC values.    108 Table 24. BIC calculations for different models with combinations of alternative specifications for the boys' data Model Polynomial order BIC calculations 1 2 3 4 5 0, 3, 0 0, 3, 1 3, 3, 0 3, 3, 1 3, 3, 3 -4054.07607 -4024.73851 -4015.11310 -4019.72073 -3993.75318  When comparing the BIC values of the competing models, it is clear that the alternative models did not improve the fit of the model to the data. Reducing the order of the trajectories failed to improve the BIC values relative to the model with all cubic-ordered trajectories. These results were suggestive of support for the 3-group, all-cubic model. Proportion of individuals in each of the groups As shown in Table 25, the second girls’ group had the greatest number of cases (N = 329), which was estimated to include 64% of the population (p < .0001). This was followed by the fourth group with 70 girls. This latter group was estimated to include 14% of the population (p < .001). Group one included the same proportion as the fourth group, although the sample size of the group was only 56 subjects (p = .062). Finally, the group with the fewest number of cases was the third group. This group had only 35 girls and was estimated to include 8% of the population (p = .0256). Group one in the boys’ data (Table 25) had the greatest number of cases (N= 350) and  109 was estimated to comprise 70% of the population (p = .000). The second and third groups had only 88 (p = .000) and 44 (p = .001) cases, respectively, and were estimated to jointly include about 30% of the population. Table 25. Probability of group membership  Girls N π* Boys N π* 1 56 .14  1 350 .70 2 329 .64  2 88 .19 3 35 .08  3 44 .11 4 70 .14 *π = probability. The calculation of posterior group membership probabilities The probability of an individual belonging to a group is referred to as the posterior probability of group membership, which is computed by using the model’s estimated coefficients (Nagin, 2005). The calculation of the posterior probabilities assigns individuals to the developmental trajectory group that best fits their behavior at each of the times of measurement. The closer the average posterior probability of assignment (AvePP) to 1, the closer is the correspondence of the model with the data. The minimum rule of thumb for an acceptable AvePP is at least .70 for all groups (Nagin, 2005). Based on the calculations of posterior group membership, I present: (a) an evaluation of model fit or adequacy based on the previously discussed criteria, (b) a description of the boys’ and girls’ models, and (c) the characteristics of the groups (i.e., the group profiles).  110 Model adequacy In the subsequent sections, I conform to Nagin’s (2005) recommendations concerning the judgment of the adequacy of the boys’ and girls’ models in their ability to identify distinct groups of BMI trajectories. Model adequacy or fit is evaluated through several diagnostic tools, including the average assignment probability for the data, the odds of correct classification, the confidence intervals for group membership probability, and the estimated group probabilities. Diagnostic 1: The average posterior probability of assignment One diagnostic of model adequacy relies on the AvePP for each trajectory group, which provides information about the correctness of the group membership classifications based on the maximum posterior probability assignment rule (Nagin, 2005). As described earlier, the closer the assignment probability is to 1, the closer is the model to the data. Further, the minimum rule of thumb for an acceptable average posterior probability (AvePP) is at least .70 for all groups (Nagin, 2005). Tables 26 and 27 report the mean posterior probabilities of membership for the individuals that were allocated to each of the groups.         111 Table 26. Average assignment probability conditional on assignment by maximum posterior probability rule for the girls’ model  Assigned group N Group 1 Group 2 Group 3 Group 4 Group 1 56 .728 .116 .0350 .121 Group 2 329 .050 .903 .029 .019 Group 3 35 .072 .099 .818 .012 Group 4 70 .125 .080 .009 .787   Table 27. Average assignment probability conditional on assignment by maximum posterior probability rule for the boys’ model  Assigned group N Group 1 Group 2 Group 3 Group 1 350 .911 .051 .039 Group 2 88 .154 .781 .046 Group 3 44 .130 .073 .797    112 The mean group posterior probability for the girls’ data ranged from .73 to .90, relatively good probabilities, while the mean group posterior probability for the boys’ data ranged from .80 to .91, again, relatively strong classification probabilities (i.e., strong correspondence of the model with the data). The results showed that all of the AvePPs were above the minimally acceptable threshold of .70, indicating ‘model adequacy.’ More specifically, the posterior probability for the 56 girls assigned to Group 1 was 0.73, while the average for the 329 individuals assigned to Group 2 was .90 (reflecting the large size of this group). For Group 3 (n = 35) and Group 4 (n = 70), the averages were .82 and .79, respectively. On the basis of the maximum posterior probability assignment rule, 350, 88, and 44 individuals from the boys’ data were assigned to Groups 1, 2, and 3, respectively. The average for Group 1 was considerably large at .91, while the average for the 88 boys assigned to Group 2 and the average for the 44 boys in Group 3 was .80. The results provided support that individuals were assigned to the group that had the largest posterior probability. Diagnostic 2: The odds of correct classification  The odds of correct classification (OCC) to a group are also based on the maximum posterior probability assignment rule. According to Nagin (2005), an OCC value that is larger than 5.0 for all the groups is indicative of high assignment accuracy of the model. Examination of the girls’ results in Table 28 shows that Groups 1 through 4 had odds of correct classification that were greater than 5, thus indicating high assignment accuracy while Group 1 in the boys’ model had an OCC that was below the recommended acceptable value (see Table 29).   113 Table 28. Odds of correct classification for the girls’ model  Group π AvePP Odds of correct classification 1 2 3 4 .14 .64 .08 .14 .73 .90 .82 .79 16.6 5.1 52.4 23.1   Table 29. Odds of correct classification for the boys’ model  Group π AvePP Odds of correct classification 1 2 3 .70 .19 .11 .91 .80 .80 4.3 17.1 32.4  Diagnostic 3: Calculating confidence intervals for the predicted point estimates of the trajectories of BMI Tables 30 and 31 summarize the mean BMIs and the 95% confidence intervals (95% CI) for each of the boys’ and girls’ trajectory groups for the four different time measures. Figures 4 and 5 show the point estimates of the BMI trajectories and their associated 95% CIs.   114  Table 30. Mean BMIs and 95% CIs by the girls’ trajectory groups  Time Predicted Mean BMI Group 1  95% CI for Mean Group 1 Predicted Mean BMI Group 2 95% CI for Mean Group 2 Predicted Mean BMI Group 3 95% CI for Mean Group 3 Predicted Mean BMI Group 4  95% CI for Mean Group 4 1 2 3 4 19.1 18.5 21.3 17.6 17.1-21.1 14.4-22.6 18.1-24.4 15.1-20.1 16.9 15.8 15.5 16.4 16.3-17.6 15.4-16.2 14.9-16.1 15.9-16.8 20.4 21.3 15.2 16.6 17.9-22.9 18.8-23.9 14.0-16.3 15.2-18.0 17.5 18.1 20.8 23.2 15.7-19.2 15.7-20.6 18.7-22.9 21.6-24.8    Figure 4. Trajectories of mean BMI and 95% confidence intervals (girls)   115 From Figure 4, across all four girls’ groups, the 95% confidence intervals overlapped for the first three time points but were more distinct towards the fourth time point. From the table, Trajectory 2 had the narrowest intervals and Trajectory Groups 1 and 3 had the widest, reflecting the large sample size of Group 2 (n = 300) and the small sample sizes of Groups 1 (n = 56 ) and 3 (n = 35). Looking closer at the table, Trajectory Groups 3 and 4 revealed narrower intervals at times 3 and 4. With the exception of Group 1, the narrowest 95% CIs were observed at time four. Table 31. Mean BMIs and 95% CIs by the boys’ trajectory groups  Time Predicted Mean BMI Group 1  95% CI for Mean Group 1 Predicted Mean BMI Group 2 95% CI for Mean Group 2 Predicted Mean BMI Group 3 95% CI for Mean Group 3 1 2 3 4 17.2 16.4 15.7 16.4 16.7-17.8 15.9-16.9 15.3-16.1 15.9-17.0 18.0 18.5 21.5 17.9 16.9-19.0 17.3-19.7 20.3-22.9 16.9-19.0 19.5 16.2 17.8 24.2 18.0-21.0 14.7-17.6 16.3-19.4 22.4-26.1    116 Figure 5. Trajectories of mean BMI and 95% confidence intervals (boys)               From Figure 5, the boys’ trajectories appeared to be more distinct than the girls’. There was more overlapping at times one and two, for all the trajectories, but more distinct appearing trajectories occurred at times three and four. More specifically, at time one, Trajectories 1 and 2 overlapped while Group 3’s trajectory was more distinct, but by time two, all three trajectories’ 95% CIs overlapped. Diagnostic 4: The estimated group probabilities versus the proportion of the sample assigned to each group The fourth diagnostic tool relies on the proportion of the sample assigned to each group on the basis of the maximum posterior assignment rule. Tables 32 and 33 report the  117 estimates of the group probabilities and the proportion of the sample assigned to each group. The results revealed an association between these estimates for both the boys’ and the girls’ data, which further supported the claim of reasonably well fitting models.  Table 32. Girls' estimated group probabilities versus the proportion of the sample assigned to the group Group AvePP π  P 1 2 3 4 .73 .90 .82 .79 .14 .64 .08 .14 56/490 = .11 329/490 = .67 35/490 = .07 70/490 = .14    Table 33. Boys' estimated group probabilities versus the proportion of the sample assigned to the group Group AvePP π P 1 2 3 .91 .80 .80 .70 .19 .11 350/482 = .73 88/482 = .18 44/482 = .09      118 A description of the groups based on Cole’s classifications for BMI For this study, BMI was used to assess the weight status of the children, wherein BMI was acknowledged to vary with age and sex. Consequently, for BMI to have interpretable value it must be compared to reference standards that take into consideration the children’s age and sex. The children’s BMIs were calculated from the reported heights and weights, and then compared to Cole’s cut-off criteria for overweight and obesity (see Table 2). Because Proc Traj provided predicted averages of BMI by time of assessment or age, for each of the trajectories, I describe the change in BMI for each of the trajectory groups by comparing the predicted BMIs with Cole’s cut-offs. I first compare the average BMI of each of the groups to the overweight cut-off value, and then I compare the predicted average BMI within the trajectory to the obesity cut-off value. I used the change in the cutoffs for obesity and overweight at the different assessment times to label the trajectories for the boys and the girls. The shaded area of Tables 34 and 35 provides the adaptation of Coles’ actual cutoffs and the un-shaded area provides the predicted average BMIs for the trajectory groups obtained through Proc Traj.            119 Table 34. Predicted average BMI for the girls  Age (months) Cole’s overweight BMI cut-off Cole’s obesity BMI cut-off Girls’ average age (months)* Group 1 predicted average BMI Group 2 predicted average BMI Group 3 predicted average BMI Group 4 predicted average BMI 24-29 30-35 18.0 17.8 19.8 19.6 30.2 19.1 16.9 20.4 17.5 48-53 54-59 17.3 17.2 19.2 19.1 52.8 18.5 15.8 21.3 18.1 72-77 78-82 17.3 17.5 19.7 20.1 78.4 21.3 15.5 15.2 20.8 96-101 18.4 21.6 100.3 17.6 16.4 16.6 23.2 *The average age for the 4 different time measures.  From Table 34, it is apparent that the Group 1 girls’ average BMI was above the cut- off for overweight at the first two measurements, obese at the third measure, and declined to within normal range BMI by age 100 months (8.5 years). Group 3 exhibited similar changes in average BMI, over time, although the decline to normal weight occurred earlier. This group encompassed girls that were obese in the first two measures, but unlike the first group, “rebounded” to within normal range BMI by age 6.5 years. The second group showed normal BMI for their age and sex throughout the 6-year span (age 2.5 to 8.5 years). On the other hand, the average BMIs for Group 4 indicated a steep increase from a normal BMI at the first measure to overweight status in the second, to obesity by age 6.5 years, which was sustained  120 at age 8.5 years. Table 35. Predicted average BMI for the boys  Age (months) Cole’s overweight BMI cut-off Cole’s obesity BMI cut-off Boys’ average age (months)* Group 1 predicted average BMI Group 2 predicted average BMI Group 3 predicted average BMI 24-29 30-36 18.4 18.1 20.1 19.8 29.9 17.2  18.0 19.5 48-53 54-59 17.6 17.5 19.3 19.3 52.5 16.4 18.5 16.2 72-77 78-82 17.6 17.7 19.8 20.2 77.7 15.7 21.6 17.8 96-101 18.4 21.6 100.2 16.4 17.9 24.2 *The average age for the 4 different time measures.  From Table 35, the Group 1 boys were found to have average BMIs that fell within a normal BMI range throughout the four cycles. The boys in this group had no indication of having been overweight or obese. On the other hand, the boys in Group 2 had a continuously changing BMI across the four measurements. They seemed to have had an average BMI within normal range at age 2.5 years, to have been overweight, on average, at age 4.5 years, becoming obese at age 6.5 years, and by age 8.5 years they had returned to normal average BMI. The third group, at time one, showed an average BMI that was overweight, a normal average BMI at time 2, overweight status at time 3, and a steeply rising average BMI  121 indicating obesity by age 8.5 years. Based on the predicted average BMIs and their corresponding cutoffs, as presented in Tables 34 and 35, I summarize the change in BMI in categorical form in Tables 36 and 37. This description is used to label the trajectories for the subsequent sections. Table 36. Classification of girls' BMI trajectories  Time Group 1  Group 2  Group 3  Group 4  1 High BMI (overweight) Normal BMI High BMI (obese) Normal BMI 2 High BMI (overweight) Normal BMI High BMI (obese) High BMI (overweight) 3 High BMI (obese) Normal BMI Normal BMI High BMI (obese) 4 Normal BMI Normal BMI Normal BMI High BMI (obese)   Table 37. Classification of boys' BMI trajectories  Time Group 1 Group 2 Group 3 1 Normal BMI Normal BMI High BMI (overweight) 2 Normal BMI High BMI (overweight) Normal BMI 3 Normal BMI High BMI (obese) High BMI (overweight) 4 Normal BMI Normal BMI High BMI (obese)   A description of the groups’ BMI trajectories The analysis yielded a 4-group model for the girls and a 3-group model for the boys,  122 both with all cubic polynomial orders. Figures 6 and 7 present the BMI trajectories of the girls and the boys. Tables 38 and 39 summarize the frequency and percentage of individuals in each of the groups for both the boys and the girls, separately. The Girls’ Data Figure 6. BMI trajectories for girls   Table 38. Descriptive statistics for the girls' model  Group Description N % Standard Error p 1 Late-declining BMI 56 14.0 7.50 0.062 2 Stable-normal  BMI 329 63.8 6.51 0.000 3 Early-declining BMI 35 8.2 3.69 0.026 4 Accelerating rise to obesity 70 14.0 4.24 0.001   123 The figure displays the BMI trajectory results of the girls’ data starting from age 24 months to 107 months. Group 1, which comprised 14% of the sample (see Table 38), contained girls that, at about the age of 30 months, had a BMI that was considered to be high, but by the age of 100 months, on average, had an average BMI that fell within normal range. Group 1 is labeled as the ‘late-declining BMI’ group. The second group, comprising about 64% of the sample is labeled as having ‘stable-normal BMI.’ A group labeled ‘early- declining BMI’ was comprised of girls who started with higher than normal BMI at age 30 months but later, at age 100 months, had an average BMI that fell within normal range. This group was estimated to consist of about 8% of the population. Finally, a fourth group, comprising 14% of the sample is described as the ‘accelerating rise to obesity’ group. At age 30 months, the girls in this group displayed within normal BMI, but by age 4.5 years their BMIs had increased to above the acceptable BMI range for their age, which was sustained as they aged to 8.5 years. Because of problems related to small sample size, the girls’ groups were reclassified by combining the early and late declining BMI groups into one group with 91 girls, which was labeled the ‘early and late declining BMI’ group.          124 The Boys’ Data Figure 7. BMI trajectories for boys     Table 39. Descriptive statistics for the boys' models  Group Description N  % Standard error p 1 Stable-normal BMI 350 70.1 5.47 0.00 2 Transient high BMI 88 19.0 4.64 0.00 3 j-curve obesity 44 10.8 3.21 0.00     125 The three-group model that best fit the boys’ data is described in Table 39. Group 1, which consisted of 70% of the sample, is labeled as the ‘stable-normal BMI’ group. The boys in this group displayed within normal average BMI throughout the four periods of measurement. A group labeled ‘transient high BM’ was comprised of boys who displayed a baseline normal average BMI, at the average age of 30 months, a steep rise in BMI at the third measure, but who had a reduction to normal average BMI by 100 months. This second group consisted of 19% of the sample. The last group, the ‘j-curve obesity’ group was comprised of boys who displayed a high average BMI at the first measure, a normal average BMI in the second, but by age 78 months was overweight and obese by age 100 months. This group comprised 11% of the sample. Potential confounders: Parents’ (surrogates’) education, age and income The bivariate associations between trajectory group membership (both for boys and girls) and the relevant potential confounders were assessed using chi-square tests for categorical variables (see Table 40 and 41). The variable, income, was collapsed into the variable presented to avoid expected cell counts of less than 5 and to allow for data release from the Research Data Centre (RDC). The data for the variable, age of the PMK, was not released from the RDC (with the exception of the p value) because the disclosure guidelines were not met due to small numbers and therefore are not included in the tables. The disclosure risk assessment ensures that no potentially confidential information about the study participants is reported. The results of the chi-square analysis indicated that membership in the girls’ BMI trajectory groups did not differ by the age of the PMK (p = .19), whereas there was a statistically significant association between the boys’ trajectory groups and the PMKs’ age (p = .005).  126 Girls’ Data The girls’ BMI trajectory group membership did not differ by the PMKs’ age (at the time of the interview) or by the household income. However, there was a statistically significant difference in the PMKs’ educational attainment across the three groups (see Table 40). One third (30.0%) of the girls who accelerated to obesity during the study period had PMKs with secondary school education or less; 21.6% and 25.3% of the girls in the stable- normal BMI group and the early-late declining BMI group, respectively, had similarly educated PMKs. The adjusted standardized residuals (AR) in the contingency table revealed that the girls with an accelerating rise in BMI to obesity had PMKs over represented in the less than secondary school education group (AR = 3.6) and underrepresented in the group with some education beyond high school (AR = -2.0). Table 40. Distribution of parents’ (surrogates’) age, income, and education and association with girls’ BMI trajectory group membership Variable  Early & late declining BMI n (%) Stable- normal BMI n (%) Accelerating rise to obesity N (%) Chi-square (df) p Household income ≤ $29,999 ≥ $30,000  22 (24.2%) 69 (75.8%)  62 (18.9%) 266 (81.1%)  18 (25.4%) 53 (74.6%) (2.24, 2) = .33 Highest level of schooling obtained       (19.65,6) = .003  127 Variable  Early & late declining BMI n (%) Stable- normal BMI n (%) Accelerating rise to obesity N (%) Chi-square (df) p < Secondary school Secondary school > High school College or university   degree (completed) 5 (5.5%) 18 (19.8%) 27 (29.7%) 41 (45.1%) 27 (8.2%) 44 (13.4%) 108 (32.8%) 150 (45.6%) 15 (21.4%) 6 (8.6%) 14 (20.0%) 35 (50%)    Boys’ Data The boys’ group membership differed significantly according to the PMKS’ age, household income, and level of education (see Table 41). The majority of the boys, in all three groups, lived in households with incomes of $30,000 or more. The ‘j-curve obesity’ group, however, had the greatest percentage (40.9%) of boys living in homes with incomes of less than $30,000 per annum; only 18.0% and 16.0% of the boys in the transient high and stable-normal BMI trajectory groups, respectively, lived in households with relatively low income. Similarly, the ‘j-curve obesity’ group had a greater proportion of PMKs (18.2%) with less than secondary school education compared with the transient high BMI or stable- normal BMI groups (5.6% and 4.3%, respectively). The adjusted standardized residuals (AR) in the contingency table revealed that the boys with j-curve obesity had PMKs over represented in the less than $29,999 household income level (AR = 4.00) and under represented in the $30,000 and more income level (AR = -4.00). Boys in the stable-normal  128 BMI group had PMKs under represented in the lower household income level (AR = -2.4) and over represented in the higher income level (AR = 2.4). The results also showed that boys in the j-curve obesity group had PMKs over represented in the less than secondary schooling group (AR = 3.7), while boys in the stable-normal BMI group had PMKs under represented in the same level of educational attainment (AR = -2.3). Table 41. Distribution of parents’ (surrogates’) age, income, and education and association with boys’ BMI trajectory group membership Variable  Stable-normal BMI n (%) Transient high BMI n (%) j-curve obesity n (%) Chi-square  (df) p Household income ≤ $29,999 ≥ $30,000  56 (16.0%) 294 (84.0%)  16 (18.0%) 73 (82.0%)  18 (40.9%) 26 (59.1%) (16.03, 2) < .001 Highest level of schooling obtained < Secondary school Secondary school > High school College or university    degree (completed)   15 (4.3%) 67 (19.2%) 92 (26.7%) 175(50.1%)   5 (5.6%) 11 (12.4%) 28 (31.5%) 45 (50.6%)   8 (18.2%) 7 (8.2%) 10 (22.7%) 19 (43.2%) (16.35,6) = .01    129 Bivariate relationships between trajectory group membership and the variables of primary interest An advantage to the use of posterior probability calculations, discussed earlier, is the development of a profile for the members of each of the identified trajectory groups. Based on these probabilities, bivariate analyses were conducted to fulfill the second and third objectives of this study (i.e., whether children’s temperament and their parents’ parenting practices were associated with the different BMI trajectories). This was accomplished by conducting analysis of variance (ANOVA) between the BMI trajectory groups for the primary variables: fussy/difficult temperament, consistency in parenting practices, ineffective parenting practices, and positive interaction parenting practices. Tables 42 and 43 provide the means, standard deviations, 95% confidence intervals (95% CIs), and the statistical significance of the continuous primary variables: the temperament and parenting variables. The analyses of variance, conducted to compare the means of the BMI trajectory groups, were conducted using SPSS 14.0. When the overall F statistic was found to be significant, Scheffé’s post hoc test was estimated. The Girls’ Data From Table 42, it can be seen that the girls’ membership in a BMI trajectory group was not associated with having PMK reports of fussy/difficult temperament or ineffective parenting. Positive interaction and consistency in parenting were associated with BMI trajectory group. The girls in the ‘accelerating rise to obesity’ group had PMKs that reported significantly higher positive interaction parenting scores compared with the PMKs of girls with stable-normal BMI (mean = 17.4 vs. 16.6). Scheffé’s post hoc test indicated that the mean difference between ‘accelerating rise to obesity’ group and ‘stable-normal BMI’ group  130 was significant at the 0.05 level. The girls who accelerated to obesity during the study period also had PMKs that reported significantly lower consistency scores compared with the PMKs of the girls with ‘stable-normal BMI’ trajectory group. Scheffé’s post hoc test was estimated and the mean of the ‘accelerating rise to obesity’ group was significantly different from the ‘stable-normal BMI’ group. Table 42. Mean parenting practices and temperament scores of girls with different BMI trajectories Variable N Mean SD 95% CI for Mean Min Max F p Fussy/difficult temperament Early & late declining BMI Stable-normal BMI Accelerating rise to obesity Total  91 329 70 490  27.1 26.9 28.5 27.2  7.7 7.9 6.6 7.7  25.5-28.8 26.1-27.8 27.0-30.1 26.5-27.9  11 9 16 9  47 55 43 55 1.27 .28 Positive interaction Early & late declining BMI Stable-normal BMI Accelerating rise to obesity Total  91 324 70 486  16.7 16.6 17.4 16.8  2.4 2.4 1.9 2.4  16.2-17.2 16.4-16.9 17.0-17.9 16.5-17.0  11 9 13 9  20 20 20 20 3.47 .03 Ineffective parenting Early & late declining BMI Stable-normal BMI Accelerating rise to obesity  91 323 70  8.9 8.7 9.5  2.9 4.2 4.5  8.3-9.5 8.3-9.2 8.4-10.5  2 1 0  18 22 20 1.04 .35  131 Variable N Mean SD 95% CI for Mean Min Max F p Total 484 8.9 4.0 8.5-9.2 0 22 Consistency Early & late declining BMI Stable-normal BMI Accelerating rise to obesity Total  90 321 70 482  13.7 14.5 13.2 14.2  3.4 3.4 3.4 3.4  13.0-14.4 14.1-14.9 12.4-14.0 13.9-14.5  5 0 5 0  20 20 20 20 5.24 .006  Boys’ Data The results in Table 43 provide the mean scores, standard deviations, and 95% confidence intervals for each of the primary variables of interest. The means of the PMKs’ scores on the fussy/difficult temperament, positive interaction, and ineffective parenting variables did not differ across the boys’ BMI trajectory groups. They did differ on the consistency score, however. The boys in the ‘j-curve obesity’ group had PMKs that reported statistically significantly, albeit marginally, lower consistency in parenting scores than did the boys in the other two BMI trajectory groups (mean = 13.9 vs. 14.3 (‘transient high BMI’) and 15.0 (‘stable-normal BMI’)). Several post hoc tests were conducted and none found a significant comparison. Accordingly, it is concluded that there is insufficient evidence to support a significant mean difference among the three trajectory groups on any of the four primary variables.   132 Table 43. Mean parenting practices and temperament scores of boys with different BMI trajectories Variable N Mean SD 95% CI for Mean Min Max F p Fussy/difficult temperament Stable-normal BMI Transient high BMI j-curve obesity Total   350 87 44 480   27.6 26.4 26.3 27.2   9.1 9.1 6.6 8.9   26.6-28.5 24.5-28.3 24.3-28.3 26.4-28.0   9 9 12 9   56 48 42 56 .86 .42 Positive interaction Stable-normal Transient high j-curve obesity Total  347 88 43 478  16.9 17.2 16.6 17.0  2.2 2.0 2.7 2.2  16.7-17.2 16.8-17.6 15.7-17.4 16.8-17.2  9 12 10 9  20 20 20 20 1.31 .27 Ineffective parenting Stable-normal BMI Transient high BMI j-curve obesity Total  346 88 43 477  9.3 9.0 8.3 9.2  3.7 3.5 3.2 3.6  8.9-9.7 8.3-9.8 7.3-9.3 8.8-9.5  0 1 0 0  23 19 15 23 1.60 .20 Consistency Stable-normal BMI Transient high BMI  344 88  15.0 14.3  3.2 3.8  14.7-15.3 13.6-15.1  4 5  20 20 3.07 .047  133 Variable N Mean SD 95% CI for Mean Min Max F p j-curve obesity Total 43 475 13.9 14.8 3.7 3.4 12.7-15.0 14.5-15.1 2 2 20 20   Discriminant function analysis Direct discriminant function analyses were conducted separately for the boys and girls; first a direct main effects model was estimated with the four primary variables as the only predictors of membership in the BMI trajectory groups and second the direct main effects model was adjusted by introducing the selected family characteristics that were identified to influence children’s weight status and parenting practices. The diagnostics supported the boys’ and girls’ models with all seven variables and therefore the results of the adjusted main effect models are presented. The objective of this exercise was to further examine the pattern of differences among the predictors, as a whole (i.e., assessing the notion of theoretical goodness-of-fit). More specifically, discriminant analysis allowed for the simultaneous assessment of temperament, parenting practices and other important variables and identified combinations that best predicted membership in the various BMI trajectory groups. Girls’ data Two relatively weak discriminant functions were extracted with eigenvalues of .06 and .01 for functions one and two, respectively; only function one was significant (F (14, 481) = 27.45, p = .02). The coefficient of canonical correlation was .23 for the first discriminant function and .07 for the second discriminant function. The two functions  134 accounted for 92% and 8%, respectively, of the between group variability. As shown in the graphic representation of the two functions, in Figure 8, the first discriminative function separated the ‘accelerating rise to obesity’ group from the other two groups. The second function failed to discriminate between the three groups. The structure matrix of the correlations between the predictors and discriminant functions, as seen in Table 44, suggested that the best predictors for distinguishing between the ‘accelerating rise to obesity’ group and the other two groups were fussy/difficult temperament, ineffective parenting, and PMKs’ education. The ‘accelerating rise to obesity’ group had higher scores, on average, for fussiness/difficulty (indicating more fussiness) (mean = 28.5, SD = 6.6) than the ‘stable-normal’ BMI group (mean = 26.6, SD = 7.6) and the ‘early-late declining’ BMI group (mean = 27.1, SD = 7.8). The ‘accelerating rise to obesity’ group also had higher ineffective parenting scores (indicating less effective parenting) (mean = 9.5, SD = 4.5) than the ‘stable-normal’ BMI group (mean = 8.7, SD = 4.2) and the ‘early- late declining’ BMI group (mean = 8.9, SD = 2.9). The educational attainment of the PMKs of the girls in the ‘accelerating rise to obesity’ group was lower than that of the PMKs of the other two girls’ groups. Table 45 shows that for the classification of the total usable sample of 481 girls, 235 (49%) were correctly classified. The ‘accelerating rise to obesity’ group was the most accurately classified group with 65% of the cases correctly assigned to their group based on the predictors.     135 Figure 8. Canonical discriminant function (girls)    Function 1 4 3210-1 -2 -3 Function 2 3 2 1 0 -1 -2 -3 -4 Accelerating Rise Stable-NormalEarly-late Declining Cases weighted by Standardized Weight Group Centroid Accelerating Rise Stable-Normal Early-late Declining Girls' BMI Trajectory Groups  136 Table 44. Results of discriminant function analysis: Structure matrix (girls)  Function 1 2 Fussy/difficult temperament Ineffective parenting Education Consistency Positive interaction Income PMK’s age .35* .28* -.26* -.61 .50 -.29 -.10 .18 .06 -.13 .62* .55* .38* .35* * Largest absolute correlation between each variable and any discriminant function.    Table 45. Results of discriminant function analysis: Classification results (girls)  Predicted Group Membership Group Early-late declining BMI Percent correct Stable-normal BMI Percent correct Accelerating rise to obesity Percent correct Early-late declining BMI 20.9%  41.5% 37.6%  Stable-normal BMI 16.3% 52.7% 31.0% Accelerating rise to obesity 12.5% 22.9% 64.6%  137 Boys’ data Two discriminant functions were extracted with eigenvalues of .06 and .02 for functions one and two, respectively. Only function one was significant (F (14, 467) = 31.77, p = .004). The coefficients of canonical correlation were .23 for the first discriminant function and .12 for the second discriminant function. The two functions accounted for 79% and 21%, respectively, of the between group variability. As shown in the graphic representation of the two functions, in Figure 9, the first discriminant function separated the ‘j-curve obesity’ group from the other two groups. The second function failed to discriminate between the three groups. The structure matrix of correlations between the predictors and discriminant functions, as seen in Table 46, suggested that the best predictors for distinguishing between the ‘j-curve obesity’ group and the other two groups were household income, PMKs’ level of education, and ineffective parenting practices. The boys in the ‘j-curve obesity’ group were more likely to reside in homes with low incomes and were more likely to have parents with lower educational attainment. Further, the PMKs of these children tended to report engaging less often in harmful and hostile parenting practices with their children. Table 47 shows that in the classification procedure for the total usable sample of 467 boys, 187 (44%) were correctly classified. The ‘transient high’ BMI group was the most accurately classified group with 46% (40) of the cases correctly classified according to the predictors.  138 Figure 9. Canonical discriminant function (boys)  Table 46. Results of discriminant function analysis: Structure matrix (boys)  Function 1 2 Income Education Ineffective parenting Consistency .74* .46* .35* .39 -.03 -.26 .23 .48* Function 1 4 3 210-1-2 -3 Function 2 3 2 1 0 -1 -2 -3 J-curve Obesity Transient-high Stable-Normal  Cases weighted by Standardized Weight Group Centroid J-curve Obesity Transient-high Stable-Normal Boys' BMI Trajectory Groups  139 Function 1 2 Positive interaction Fussy/difficult temperament PMK’s age .21 .14 .06 -.47* .42* .20* * Largest absolute correlation between each variable and any discriminant function.   Table 47. Results of discriminant function analysis: Classification results (boys)  Predicted Group Membership Group Stable-normal BMI Percent correct Transient high BMI Percent correct j-curve BMI Percent correct Stable-normal BMI 42.7% 33.4% 23.9% Transient high BMI 32.6% 46.2% 21.3% j-curve BMI 26.8% 28.6% 44.6%    A brief summary of the major results The various procedures and data analyses used to study the different BMI trajectories of children aged 2-8 years and their risk factors (fussy/difficult temperament and parents’ parenting practices) revealed the following: • Group-based mixture modeling analyses were conducted to identify the number and types of distinct trajectories in the development of obesity in a representative sample  140 of Canadian children who were between 24 to 35 months of age, at baseline, and followed biennially over a 6-year span. o The analyses established three different BMI trajectories for the boys, namely: stable-normal BMI, transient-high BMI, and j-curve obesity. The analyses revealed four different trajectories of BMI for the girls: stable-normal BMI, early-declining BMI, late-declining BMI, and accelerating rise to obesity. Because of small sample sizes, the early and late declining groups were merged into one group, labeled early-late declining BMI. • A bivariate analysis, with ANOVA, was used to determine the relationship between children’s temperament and the development and maintenance of obesity (trajectory of BMI); the results indicated that there was no association between the BMI trajectories and reported fussy/difficult temperament for the boys and the girls. • Bivariate analyses, with ANOVA, were conducted to examine the associations between the gender-specific BMI trajectories and the different parenting practices. o The results indicated that, compared with girls in the stable-normal BMI and the early-late declining BMI groups, girls who belonged to the accelerating rise to obesity group had parents who reported greater frequency of positive encounters such as praising, laughing and spending quality time with their children. Girls in this latter BMI trajectory group also had parents who reported being inconsistent in their discipline. o The analyses failed to detect statistically significant associations between parenting practices and the BMI trajectory groups for the boys. • Discriminant analysis was conducted to assess the theoretical notion of goodness-of-  141 fit between parenting practices, children’s temperament, their family environment and the BMI trajectory groups. o The girls’ model revealed that when a girl was described to have a fussy/difficult temperament and had parents with ineffective parenting practices (i.e., critical and disapproving) and low educational attainment, she was more likely to be classified in the accelerating rise to obesity group. o The analyses revealed that for the boys, the best simultaneous predictors of j- curve obesity were low scores on ineffective parenting practices, household income, and educational attainment. That is, when a boy resided with parents who scored low in ineffective parenting and had low socioeconomic status (i.e., lower household income and education) he was more likely to be classified in the j-curve obesity group. The theoretical notion of goodness-of- fit was not supported by the boys’ data.  142 Chapter 5: Discussion  This chapter presents the principal conclusions emerging from this study. The chapter is divided into five main sections: (a) an overview of the study objectives and procedures, (b) a summary of the results, (c) an attempt to link the findings with current and pertinent literature, (d) a discussion of gender variability in children’s temperament and parenting practices in the context of obesity, (e) study limitations and future implications, and (f) a chapter summary. General overview  The major objectives of this study were to identify age-related trajectories of change in BMI and to investigate the association between these developmental trajectories with children’s temperament, particularly the extent to which they are fussy or difficult and their parents’ parenting practices (i.e., ineffective parenting, positive parenting, and consistency in discipline). The findings presented here can inform the children’s health field about the relationships between temperament, parenting and the development of childhood obesity. Specifically, this study is one of the first to examine the degree to which parenting practices, in interaction with the extent to which children have fussy/difficult temperaments, predict trajectories in growth that lead to or perpetuate obesity. And, the study is the first to establish longitudinal links between children’s temperament, parenting practices and childhood obesity.  The study was divided into three phases. The first phase, which applied SAS® Proc Traj, identified the developmental trajectories of BMI of the boys and girls, separately. The  143 second phase determined whether there were significant associations between parenting practices and temperament and the different BMI trajectories. The final stage identified significant predictors of the different trajectories in the context of the goodness-of-fit between the children’s temperaments and their parents’ parenting practices. The statistical techniques used in phases two and three were ANOVA, to examine the bivariate relationships, and discriminant analysis, to explore the fit between the children’s temperaments and their parents’ parenting practices. This study used data from Cycles 2 through 5 of the Canadian National Longitudinal Survey of Children and Youth (NLSCY), specifically those of children who were between the ages of 24 to 35 months, at baseline (Cycle 2), and followed the children, biennially, until they had reached the ages of 96 to 107 months (Cycle 5). The sample was selected following specific criteria. Firstly, children missing more than two BMI measures, in the dataset, between the ages of 2 and 8.9 years were excluded to ensure that the analyses were based on a minimum of three measures of BMI. Prior to executing this requirement, all cases with identified BMI outliers were removed. The excluded BMI values were presumed to be biologically implausible, meaning that they were the result of error in data management or reporting rather than reflective of actual weights and heights. Because the weights and heights were reported by the person most knowledgeable about the child (PMK) and likely to be subject to significant error, the outlier exclusion criterion added a level of rigor to the analyses.  The project was designed to address several other limitations noted in past research. First, the study included both children’s temperament and their parents’ parenting practices to incorporate the notion of goodness-of-fit. Second, the fitting of a discrete mixture model to  144 longitudinal data allowed for the identification of distinct trajectories of change in BMI. Third, the sample was diverse in terms of the children’s familial socioeconomic status so it is relatively more representative of the general population than have been many of the samples studied in the past. Decisions made to strengthen some aspects of the design resulted in limitations in other respects (e.g., generalizability). Because of the exclusion criteria, the number of children in the cohort was reduced from a possible 1,890 cases to 972 (490 girls and 482 boys). The excluded children differed significantly in important ways from those included in the study; namely, the persons most knowledgeable about the children, the PMKs, differed in age, educational attainment, and household income. However, the PMKs’ reports of their children’s temperaments, and the extent of their positive interactions and ineffective parenting did not differ. This suggests that although the demographics of the two cohorts differed significantly, the PMKs’ backgrounds did not influence their reports with the exception of their responses to questions about consistency in discipline. The excluded children’s PMKs reported greater inconsistency in their use of discipline. The cohort was small compared with the number of children included in the NLSCY because of the restriction in age (24-35 months) at time one (Cycle 2). However, the selected sample provided some noteworthy advantages. One advantage was the selection of an age group young enough that the temperamental dimensions were measured with less concern that the measure was confounded by behavioral or social modification (i.e., the children were not old enough to have been exposed to factors that could have influenced their personality development). Second, the age group studied covered an age span when most parents initiate and enforce discipline over their children. Developmental studies have shown that this age  145 range is also significant in predicting BMI at a later age. Response bias is another potential factor that could have affected the conclusions of this study. There are multiple factors that could have influenced the PMKs’ reports of their children’s temperament; parent-reported temperament scores may reflect parental expectations, parental psychological well being (stress or depression), and their own characteristics. For example, depressed or stressed mothers tend to report more difficult temperament than do non-depressed mothers of infants and toddlers (Austin et al., 2005; Mantymaa et al., 2006; Warren & Simmens, 2000). Summary of the findings The following sections provide a brief summary of the findings based on the four research objectives. A summary of the prevalence of overweight and obesity for boys and girls is first described before the specific findings, with respect to the research questions, are tackled. Prevalence of overweight and obesity The overall prevalence rates of overweight and obesity for the children in the study cohort were at their highest at the first measure, when the children were aged 24-35 months, and declined considerably (especially for obesity) by the last measure (ages 96 to 107 months). The prevalence rates for obesity were higher for the girls, throughout the four cycles. The greatest prevalence rate of overweight was observed for the girls during Cycle 3 (48 to 59 months). There was a decline in the prevalence rate of obesity for the boys and girls by the last measure (ages 96 to 107 months). The rates of overweight for the boys remained fairly constant over time and were lower than the rates of obesity in the first three cycles. The  146 lowest rate of obesity was observed for boys during the fifth and last cycle. Research question 1  What are the number and types of distinct trajectories in the development of obesity in a representative sample of Canadian children who are between 24 to 35 months of age, at baseline, and followed biennially over a 6-year span? The current study identified multiple trajectories of BMI for the girls and boys during the age span of 2 to 8 years. The analyses revealed a trajectory in which the boys and girls never were obese; these children were grouped into a ‘stable-normal BMI’ group. The analyses also identified two girls’ groups and one boys’ group for whom the trajectories displayed fluctuations in BMI, over time, but in which the BMI, on average, fell within acceptable or normal limits at the last measure. The magnitude of the average BMI, at the final measurement for these two girls’ trajectory groups, was similar and therefore the two groups were combined to form an ‘early-late declining BMI’ group; the boys’ group was labeled the ‘transient high BMI’ group. Finally, the analyses identified a group of obese children. There were some boys who were obese at 2 years of age and who had extremely high BMIs by the last measure, although some reduction in BMI was seen at the time of the second and third measurements; this group was labeled the ‘j-curve obesity’ group. The ‘accelerating to obesity’ girls’ group exhibited normal BMIs, on average, at 30 months of age, and had BMIs that increased steeply to obesity by the last measure.    147 Research question 2 What is the relationship between children’s temperament and the development and maintenance of obesity (BMI trajectory group membership)? The second objective of this study was to determine whether there is an association between children’s temperaments and the trajectory groups of BMI. That is, the study investigated whether there was an association between membership in a particular BMI trajectory group, specifically one associated with obesity, and being reported to have a ‘fussy/difficult’ temperament. The results of the bivariate analyses indicated that membership in a BMI trajectory group, for both girls and boys, was not associated with having reports of a fussy/difficult temperament. Research question 3 What is the relationship between different parenting practices and the development and maintenance of obesity (BMI trajectory group membership)? The third objective of this study, using bivariate analyses, was to determine whether there is an association between children’s parents’ parenting practices (i.e., consistency in discipline, positive interactions, and ineffective practices) and the trajectory groups of BMI. Among the three parenting practices examined, consistency in discipline and positive interactions were found to be associated with the girls’ trajectory that led to obesity. PMKs who reported inconsistency in the use of discipline and the use of positive interaction were more likely to have girls classified in the ‘accelerating to obesity’ trajectory group compared with the other trajectory groups. There were no significant mean differences in the measures of consistency in discipline or positive interaction to support an association between the  148 boys’ trajectory group membership. The analyses failed to identify ineffective parenting as a correlate of BMI trajectory for either the boys or the girls. Research question 4 Are temperament and parenting practices independent predictive factors or synergistic in the development of childhood obesity? This study also assessed whether the theoretical notion of goodness-of-fit between early childhood temperament and parents’ parenting practices is related to the different trajectories of BMI (i.e., whether temperament and parenting practices are predictive factors or synergistic in the development of childhood obesity). Specifically, the objective was to investigate, via discriminant analysis, whether children’s changes in BMI differ when their parents’ parenting practices are dissonant or consonant, particularly in children with fussy/difficult temperaments. The structure matrix of the correlations among the predictors examined and the discriminant functions produced by the discriminant analysis of the girls’ data suggested that the strongest combined predictors of membership in the ‘accelerating to obesity’ group, rather than the other trajectory groups were having a fussy/difficult temperament, ineffective parenting, and relatively low parental education. While the multivariate analyses failed to identify consistency in discipline and positive interactions as risk factors for the development of obesity, ineffective parenting practices distinguished the unhealthful from the healthful BMI trajectories for girls. Relatively higher scores of ineffective parenting practices, when the children were 2 years of age, were predictive of membership in the ‘accelerating to obesity’ trajectory group. In addition, reports of a fussy/difficult temperament, at age 2, were found to be predictive of girls’ membership in the ‘accelerating to obesity’ trajectory group.  149 Thus, the parents of these young girls with putatively challenging temperaments would be characterized as highly critical and disapproving. The corresponding structure matrix of correlations among the predictors and discriminant functions for the boys’ data suggested that the best predictors for distinguishing between the ‘j-curve obesity’ group and the other two groups were household income, parental education, and ineffective parenting practices. The boys in the ‘j-curve obesity’ group were more likely to reside in homes with low incomes and were more likely to have parents with lower educational attainment. Somewhat unexpectedly, the parents of these boys tended to report less engagement in harmful or hostile parenting practices compared with the reports of other parents. The theoretical notion of goodness-of-fit (between early childhood temperament and parents’ parenting practices) was not supported by the boys’ data. Connections and parallels with the existing literature In the following sections, parallels between the major findings of this study and the available literature are illustrated. A brief comparison of the prevalence of overweight and obesity among the study cohort and national prevalence rates is first provided. Finally, gender variability in parenting and childhood temperament is discussed. Prevalence of overweight and obesity The cross-sectional prevalence rates of overweight and obesity among the 2-3 and 8-9 year-old children in this study are consistent with those reported by Canning et al. (2004) and Statistics Canada (2002); the rate of obesity or overweight among the boys and girls is highest at ages 2-3 years and lowest at ages 8-9 years. About 45% of 2-3 year olds and about 27% of 8-9 year olds were classified as obese or overweight in the first cycle of the NLSCY  150 while the prevalence rates for this study’s subset of children were 37.2% (14.8% overweight and 22.4% obese) for 2-3 year-old boys and girls and 22% (13.3% overweight and 8.7% obese) for 8-9 year-old boys and girls. The lower prevalence rates of overweight and obesity, in comparison with the larger sample is likely due to the elimination of the weight and height outliers (i.e., PMKs’ reports that resulted in biologically implausible BMIs). The prevalence rate of overweight/obesity for this study’s children, aged 2-3 years, was higher for the girls (42.2%) than for the boys (32.1%), which is consistent with the preschool aged children’s status in Ogden et al.’s (1997) and Magarey et al.’s (2001) studies, which used Cole’s (2000) classification criteria. The higher rate of obesity among girls is also consistent with Martorell et al.’s (2000) study that examined the prevalence of overweight and obesity in preschool children from 50 developing countries. The higher overall prevalence of overweight/obesity among girls, across the four cycles, compared with boys, contradicts many other studies wherein the prevalence of obesity has been found to be higher in boys than in girls (Gauthier et al., 2000; Kromeyer- Hauschild et al., 1999; Mustillo et al., 2002; Thorpe et al., 2004; Troiano & Flegal, 1998). It is important to note that the age groups and classification criteria are not identical across studies. More specifically, Statistics Canada (2002) results indicated that, in addition to the overall increase in the relative frequency of overweight and obese children, since 1994, more boys than girls are classified as overweight and obese. For example, in the 1998/1999 NLSCY cohorts (third cycle), an estimated 53% of boys were classified as overweight or obese (38% overweight and 19% obese) as opposed to 52.1% (35% overweight and 17% obese) of girls. Willms, Tremblay, and Katzmarzky (2003) studied Canadian boys and girls aged 7 to 13 years who participated in the NLSCY and the 1981 Canada Fitness Survey and  151 concluded that the prevalence of overweight is higher among boys. As seen here, whether there is a gender difference in the prevalence of obesity is generally unclear; the equivocal findings may be attributed to difference in the measurement of obesity (e.g., skin fold thickness versus BMI), in the various age groups studied, or in the reference norms used to categorize children (Livingstone, 2000). In addition to these methodological differences, the findings in this particular study may be potentially biased because of the exclusion of a significant number of children with missing data, a reliance on parental reports of the children’s weight and height, and exclusion of biologically implausible values. Further, gender variations in the prevalence of overweight and obesity may differ by ethnicity/race or other demographic, geographic, sociocultural or health factors. Significant differences have been observed in the prevalence of overweight by gender and ethnicity/race. For example, the prevalence of overweight is highest among Mexican-American girls and lowest in non-Hispanic white boys (between 2-5 years of age) (Ogden et al., 1997). Furthermore, it is well established that predictors of high BMI among children are: low levels of parental education and income. There is gender variability found within these risk factors, as well. There are more obese boys than girls in the lower socioeconomic context while more girls than boys are obese within the higher socioeconomic strata of the population (Danielzik et al., 2004; Garn & Clark, 1975; O’Dea & Caputi, 2001). Finally, Willms et al. (2003), using the NLSCY, also concluded that the prevalence of overweight and obesity among children tends to decrease with age, which is consistent with the findings of the current study. BMI trajectories Utilizing the mixed modeling approach in this study resulted in the identification of  152 subgroups of boys and girls with distinct levels of BMI and has provided average patterns of change over a 6-year span for the subgroups. It is pertinent to acknowledge that the results derived from applying the group-based methodology are approximations of population differences in developmental trajectories, and that the BMI trajectories presented in this study are based on group means over a specific time period. According to Nagin (2005), “The population is neither comprised of literally distinct groups nor are the parameters describing the population variation normally distributed…the models54 are just approximations of a more complex reality” (p. 45). Direct comparison of the results of this study and others is very difficult because of differences in the ages at which the values of height and weight were reported or measured, the measures of adiposity taken, and the reference cut-offs used to categorize the children’s adiposity. Most importantly, this study explored the variation in BMI in clustered and distinct groups, whereas the majority of studies used correlational methods to track changes in BMI. Nonetheless, the strength of applying Proc Traj to identify distinct BMI trajectories has been valuable in terms of understanding the different ways in which a child may develop obesity. Compared with other published results of the NLSCY (Statistics Canada, 2002), which studied the same children, who were aged 2-11 at Cycle 1 and 6-15 years in Cycle 3, the prevalence of children that were never overweight (the stable-normal BMI groups) is higher in this study cohort than in the national sample as a whole. According to Statistics Canada, approximately 44% of the children were never overweight over the 4-year span, while 70% of the boys and about 64% of the girls in the current study were classified to be in the stable- BMI trajectories. Of the children studied in the Statistics Canada report, about 10% were  54 The models are growth curves and group-based models.  153 consistently overweight, which approximates the percentage of boys classified in the ‘j-curve obesity’ group (11%). Only one other research study has explored the variation in BMI in clustered and distinct groups, although the analysis was based on a dichotomous model because the dependent variable was binary: obese versus non-obese, rather than continuous, as employed in the current study. Consequently, there is a limit to the comparison that can be made between the results of the studies. Nonetheless, this other longitudinal study, by Mustillo et al. (2003), also reported the presence of groups of children with normal and declining BMIs. The pattern of change in the boys’ BMI classified in the ‘j-curve obesity’ group may be consistent with the occurrence of rebound obesity reported by various researchers. ‘Adiposity rebound’ has been defined as a gradual increase in BMI after having reached a minimum value, which on average occurs between the ages of 5-7 years for children in the 50th percentile and earlier for children in the 95th percentile of BMI (Dietz, 2000). Early rebound occurs at about 4.8 years (Whitaker et al., 1998). This pattern corresponds well with the graphic presentations of the boys’ trajectories presented here where there is a noticeable decline in BMI at the second measure in the ‘j-curve BMI’ group (average age is 53 months (4.4 years)) and then a gradual increase in BMI (observed at the third measure when the boy were 6.4 years, on average).55 The occurrence of ‘adiposity rebound’ is related to an increased risk for the development of childhood obesity that persists into adolescence (Diez, 2000; Siervogel et al., 1991) and possibly into adulthood (Rolland-Cachera et al., 1987; Whitaker et al., 1998).56 Dietz (2000), however, suggested that BMI at the time of rebound is  55 This interpretation is based solely on the graphic presentation of change in BMI and the corresponding average BMI at the relevant points in time.   154 a stronger predictor of later BMI than the actual timing of the rebound. For example, children with high BMI, at rebound, are more likely to be overweight or obese as adults. In summary, the study findings presented here suggest that there are previously undocumented BMI ‘developmental’ patterns that may be important for policy and public health decisions. It appears that children have different pathways to obesity and at different ages and, in particular, there is a group of children that may be described as being of normal weight at one time and who subsequently appear obese; these children may be overlooked by one-time assessment and intervention programs if such interventions are not timed appropriately. Temperament and BMI trajectories: Bivariate associations No associations were found between children’s temperament and obesity – a primary hypothesis at the outset of this work. The measure of a fussy/difficult temperament between 2-3 years of age was not found to be associated with any of the BMI trajectory groups (for boys and girls). The findings are consistent with Kramer et al.’s (1986) analyses of a prospective cohort of 351 healthy infants followed from birth to 24 months and contradict the majority of previous studies, both cross-sectional and longitudinal, that have established an association between the presence of a difficult temperament and the development of obesity in infants and children (Agras et al., 2004; Carey, 1985; Carey et al., 1988; Wells et al., 1997). Even with the use of a variety of measures of temperament and diverse definitions and classifications of obesity, a shared finding across these previous studies is that a relationship exists between temperament and obesity. However, the features of these studies  56 Children who experience earlier adiposity rebound usually have above average BMIs.  155 that limit the possibility of comparison and generalizability are many. First, the majority of these studies were based on small sample sizes and on samples that were not representative of the general population (i.e., the participants were described as mainly white and of middle- income families with a high percentages of them having parents with college education). Second, some of the studies assessed temperament at a later age (4-5 years old) thus limiting their ability to distinguish between actual temperament characteristics or emerging behavior as the result of the children’s interactions within their larger social environments. Kremer et al. assessed temperament during infancy (at 2 weeks of age) and found no relationship between temperament and subsequent overweight, which could, very significantly, be attributed to the stability of temperament during young age. Consequently, it is unclear as to whether the equivocal findings are the result of issues of construct validity (whether temperament is actually being measured when children are over 2-3 years of age) or potentially biased cohorts. Parenting and BMI trajectories: Bivariate associations A considerable number of studies have documented relationships between positive interactions between parents and children (measured as warmth, nurturing or responsiveness) and consistency in discipline, and children’s and adolescents’ well being (Baumrind, 1967; Brand et al., 1990; Frick et al., 1999; Pettit & Bates, 1989;). Maternal warmth was described by Bates et al. (1982) as the most consistent correlate of children’s competence. Positive encounters, through showing praise and approval, promote conscience development, self esteem, autonomy, and prosocial behavior (Barber, 1997; Hoffman, 1970; Zhou et al., 2002). Positive interactions are integral to, and a protective component of, the well being of children if they are coupled with clear and consistent expectations, especially for children in the 2-3  156 year age range. Based on Maccoby and Martin (1983) and Barber and Olson’s (1997) work, an optimal parenting environment is one that combines positive emotional interactions, firmness, and consistency in establishing and enforcing guidelines. Toddlers are at an age when self-regulation (emotional and biological) is just emerging, and for parents to act as positive agents in their children’s adjustment, they must contribute to that regulation and modulation of behavior through being both warm and consistent in their approach. The mixed results of the bivariate analyses of the girls’ data in the current study (PMKs that self-reported being warm and inconsistent in their discipline had greater likelihood of having girls in the ‘accelerating to obesity’ group) 57 add to the complexity of the already conflicting literature on the relationship between parenting practices and childhood obesity. This study is one of a few studies to find associations between two opposing practices (i.e., warm interactions and inconsistent parenting practices) and obesity. The findings support the notion that parents apply different parenting strategies with their children, and the combination of inconsistency in discipline joined with warm interactions is not entirely exceptional in the field of obesity and feeding practices. Chen and Kennedy (2004) reported that the parents of 8-10 year-old obese Chinese children applied democratic parenting styles and showed permissive practices in relation to food intake. In their study, a more democratic parenting style was associated with higher amounts of food consumption including sugary foods. The girls of this study classified in the ‘accelerating to obesity’ group had parents who were both inconsistent in the discipline they proffered and warm in their interactions  57 One needs to be careful in interpreting the results of the bivariate analysis because there is no evidence to indicate that the children with inconsistent parents (in the application of discipline) also had warm parents.  157 with their children (according to their self reports). The combination of the two different parenting practices could provide an environment conducive to the development of obesity through the creation of a highly permissive attitude. Warmth is an integral protective component of the well being of children if it is complemented with clear and consistent expectations; these parenting practices can then provide an environment conducive to healthful behavior and proper self-regulation. Consistent with the idea that the parents of the girls in the ‘accelerating to obesity’ trajectory may have parented in a relatively permissive manner, the results of this study may be comparable to the conclusions drawn by Brewis (2003). In Brewis’s study, children (6-12 years old) were more likely to be obese if they had parents who were relatively more permissive or less authoritarian. Unfortunately, the report of the study did not include a comprehensive description of the tools used to measure the various parenting styles and authoritative parenting was not evaluated. It seems that permissiveness and authoritarian styles were conceptualized as two extremes on one continuum rather than two distinct dimensions. Another study, by Rhee et al. (2006), examined the relationship between parenting styles and obesity in a national sample of 872 children aged 54 months and their parents. The mothers were classified as authoritarian, authoritative, permissive, or neglectful according to their sensitivity to their children’s needs and expectations for self-control. The results showed that, relative to an authoritative parenting style, mothers with an authoritarian, permissive, and neglectful parenting style (when their children were aged 4 years) were more likely to have children who were overweight two years hence (aged 6 years). A major limitation of the study, however, is the lack of adequate validity evidence for the parenting style measures used. The researchers did not apply a standardized measure of parenting style and acknowledged the need for further  158 validation against other models of parenting styles. Agras et al. (2004) failed to detect a significant relationship between parenting styles (authoritarian, authoritative, and permissive styles of behavior) and overweight and obesity in a cohort of 150 children followed from birth to 9.5 years. The authors, however, reported high attrition among the participants and those who dropped out differed significantly in education from those who continued. Further, the study did not provide information about the age at which parenting was measured or if repeated measures were obtained. Mustillo et al. (2003) were also unsuccessful in establishing relationships between different trajectories of childhood obesity (based on group-based memberships) and harsh or overprotective parenting, lax supervision, and the development of obesity. It is perhaps important however that the children in their study were older than 9 years of age and resided in rural areas. Furthermore, they used a dichotomized form of the dependent variable (obese versus not obese), which may have resulted in a loss of information and higher rates of misclassification. On the other hand, Strauss and Knight (1999) also reported no association between parental support and the development of obesity. In their study, children who became obese were just as likely to be hugged, kissed or spanked as were children who did not develop obesity. Further, Gable and Lutz (2000) were unable to predict obesity in children 3-10 years of age on the basis of their parents’ parenting styles (authoritative vs. authoritarian). In another prospective study, Lissau-Lund and Sфrensen (1992) reported conclusions that contradicted these findings: families of obese children were noted to have a tendency to be negligent or to show less compassion. Few researchers have examined consistency in parenting in the context of obesity. The finding here that the application of inconsistent discipline is associated with an obesity  159 trajectory for girls is consistent with Decaluwe et al.’s (2006) study, which attempted to relate parental behavior to psychological problems among 10-16 year-old obese youth. They reported that a large percentage (about 40%) of the parents of children with problematic weight scored high on inconsistency in discipline (they showed less initiative to correct misbehavior and were less consistent in punishing). Although the age gap between this current cohort and Decaluwe et al.’s participants is fairly wide, consistency in discipline has been shown to not vary with children’s age (Chao & Willms, 2002). Home environments that are inconsistent in discipline most probably lack effective rules (both enforcing and adhering to rules) and provide unstructured environments. Unstructured home environments presumably lack structured eating habits and behavior that may be protective against obesity (e.g., having regular meal times or frequently eating together as a family). Inconsistent practices could extend to other aspects of children’s lives such as in shaping their level of activity and consequently their weight. For example, Gentile and Walsh (2002) reported that children watched less television and engaged in more alternative activities when their parents applied consistent discipline (e.g., putting limits on the amount of time television watched). More distinctively, children 4-13 years of age that participated in the first cycle of the Canadian National Survey of Children and Youth (NLSCY, 1994), and who resided with parents with consistent, non-punitive or non-aversive parenting styles, were shown to be involved in higher levels of physical activity (Cragg, Cameron, & Russell, 1999). Similarly, children residing with parents who endorse authoritative (warm and demanding) parenting practices are more active than are children of parents with authoritarian (lack warmth and highly demanding) parenting styles (Gable & Lutz, 2000).58  58 Unstructured family environments also have been related to eating disorders in adolescent  160 Theoretical goodness-of-fit Children’s temperament, parents’ parenting practices, and the prediction of BMI trajectory group membership This study is the first to document temperament and parenting practices as interactive factors in the prediction of the development of obesity in girls. The results of the multivariate analyses suggest that ineffective parenting practices and a fussy/difficult temperament differentiate girls’ BMI trajectories (i.e., fussy/difficult 2-year-old girls, whose parents were disapproving and highly critical, displayed an accelerating type of growth that was classified as obesity by age 8 years). If parenting practices can be viewed as indicators of parents’ feeding practices, we may be able to explain this important finding. Parents of fussy and irritable children exhibit higher levels of food prompting as methods to control tantrums (Agras et al., 2004; deVries, 1984). And, fussy and highly emotional children tend to exhibit persistent tantrums over food (Agras et al., 2004). The PMKs of the fussy girls, in the present study, may have used food prompting as a means of controlling the tantrums arising from the girls’ temperament. Parents often find it challenging to differentiate between tantrums related to food and those associated with the emotional aspects of a child’s temperament. The interaction between the child’s temperament and the parent’s response may create feedback cycles leading to both parental overfeeding and children’s overeating, which may persist over time and lead to greater degrees of overweight and obesity. The test of the goodness-of-fit model that analyzed the associations between  girls, and bulimia in adolescents has been related to obesity during childhood. In this context, obesity in a chaotic household is related to adolescents’ inability to regulate their own food intake.   161 parenting, temperament, and changes in BMI, and that was confirmed for the girls, is consistent with previous studies that examined the fit between temperament and an array of parenting practices in the context of children’s adjustment.59 For example, Morris et al. (2002) suggested that children (of about 7 years of age) who are highly irritable are at higher risk for developing externalized behavior problems when parents are hostile (i.e., have overt negative affect with demonstrations of anger with yelling and spanking), while parental psychological control (covert hostility, intrusiveness, and manipulation of the child’s feelings) places the children at risk for internalizing problems. Similar outcomes have been reported when children with difficult temperaments interact with parents who are harsh, rejecting, lacking in warmth, intrusive, inconsistent in disciplinary approaches, or generally restrictive (Belsky et al., 1998; Maziade et al., 1990; Rubin et al., 1998; Vitaro et al., 2006). More pertinently, Gilliom et al. (2002), in their longitudinal study, found that a poor fit between a difficult temperament (measured in toddlerhood) and parenting (described as high negative control and low in warmth) resulted in ineffective self-regulation 2 years later (the children were incapable of waiting to open a gift). Gilliom et al.’s study provides insights that could explain the process by which the contribution of the interactive factors studied here (temperament and parenting) could lead to girls’ obesity. Children who are inclined to react negatively may be more reactive to a critical and disapproving parent and more vulnerable to the impact of their parenting practices. Irritable and fussy children that must interact with critical parents are at risk for high levels of conflict  59 Some researchers have hypothesized that children’s temperament is a moderator of the relationship between parenting practices and children’s adjustment (Morris et al., 2002) while others have focused on parenting as the moderator (Bates et al., 1998; Gilliom, 2002; Rubin et al., 1998).   162 and negative parent-child relationship processes (arising from poor fit between the parenting practices and the children’s temperament).60 These recurrent negative encounters could contribute to the development of maladaptive behavior (i.e., lack of self regulation), which could extend to a lack of regulation over food consumption. Subsequently, food may possibly be used as a comfort measure even in the absence of hunger, which is a phenomenon that has been observed in girls and is related to obesity (Birch et al., 2003). Confounding variables Overall, the significant confounding variables included in the analyses of the current study -- which were not central to the goals of the study -- played a role in the development of obesity among the boys and are consistent with those described in the literature. Low income and low educational attainment have been described as risk factors for preschoolers, school-age children, and adolescents’ obesity in many studies in Canada and the US (Johnson-Down et al., 1997; Mei et al., 1998; Wang, 2001; Weicha & Casey, 1994). This is usually attributed to the consumption of high carbohydrate foods, lack of exercise and engagement in sedentary behavior (Casey et al., 2001). For example, more television watching has been observed among children from lower educational attainment and income homes (Bernard-Bonnin et al., 1991; Casey et al., 2001; Kimm et al., 1996). Some beliefs reported by lower socioeconomic parents may affect their feeding practices and may also lead to obesity in children, including the belief that heavier infants are healthier, that infants are not satiated by breast milk or formula alone (which results in the early introduction of  60 A considerable body of research has documented associations between ineffective parenting and a range of childhood outcomes. For example, Richman, Stevenson, and Graham (1982) reported four times more problematic behavior among children addressed critically and disapprovingly by their parents.   163 solids or the addition of cereals to bottles), and the use of food as a tool to shape behavior (Baughcum et al., 1998).61 Confidence in the findings The ability to predict the course of a girl’s growth and potential for obesity from her parent’s parenting practices and her temperament, measured during the first 2-3 years of life, is suggestive of the relative stability of the measures. For the boys, it is possible that interactive effects between temperament and parenting could unfold at a later time. Alternatively, the inability to find an association may be related to the different socialization experiences of boys or simply to having had a biased sample. The small size of the obesity trajectory group may have resulted in smaller numbers of children with high scores on the temperament variable and which may have subsequently reduced the statistical power of the analyses and resulted in a Type II error. Further, the boys classified in the obesity group were more likely to have resided in homes with relatively low income and educational attainment, which are strong risk factors for obesity and possibly strong confounders of parenting practices and parents’ reports about their sons’ temperaments. Gender differences Gender variability in temperament and obesity  There is general consensus that there is little variability in temperament between boys and girls (Chess & Thomas, 1984; Rothbart, 1986); more specifically, gender variability in temperament is minimal in infancy (Rothbart, 1987), although some differences become  61 The study focused on the child-feeding practices of lower socioeconomic parents. This does not exclude the possibility that the same beliefs are also embraced by parents of higher socioeconomic status.  164 apparent in toddlerhood (Kohnstamm, 1989). In their review of the literature on the subject of gender variability in temperament and behavioral outcomes, Guerin, Gottfried, and Thomas (1997) suggested that researchers have been inconsistent in their conclusions and different researchers have reported different patterns of relationship between temperament and behavior problems for boys and girls. The relationship between the initial variability in temperament and subsequent variability in children’s development tends to be moderated by gender (Wachs & Kohnstamm, 2000). This could be due, in part, to the different contextual settings in which boys and girls reside (e.g., parental reactivity to gender-characteristic behavioral patterns) (Wachs & Kohnstamm, 2000). Group differences emerged in the current study, suggesting that there may be a gender-specific relationship between temperament and the development of obesity. Surprisingly little research has addressed this issue in the field of temperament studies and specifically in the examination of temperament and obesity. In the review of the literature, there was no reference made to possible gender differences in any of the studies on the relationship between temperament and obesity (Agras et al., 2004; Carey, 1985; Carey et al., 1988; Wells et al., 1997), and some studies, including Anderson et al.’s (2004), enrolled girls only. It is important to note that the gender variability found here could possibly be explained by a selection bias generated by the strict inclusion criteria. There may not have been enough boys who scored high on the temperament scale and consequently diminished the power to detect statistical significance. According to Thomas and Chess (1977), only 10% of the population is likely to report difficult temperament in children. Schaffer (1996) reported that the relatively global characteristics of temperament (i.e.,  165 whether a child is characterized as easy, difficult, or slow to warm up) show greater stability and predictability than do various singular dimensions. Compared with other studies, this study focused on one dimension of temperament, which could explain why no relationship was found between temperament and obesity for the boys. Only the fussy/difficult emotional aspect of temperament was examined here while most researchers have used the clusters of dimensions that group children into temperamental profiles. Nevertheless, the most predictive and stable dimension of temperament is believed to be the emotionality aspect, which is what was considered here. Gender variability in parenting and obesity The issue of whether different parenting practices are directed toward boys and girls has been debated extensively; the earliest literature suggested that there were no gender differences in the process of parental socialization of children (reviewed by Huston, 1983). More recently, in studying factors that may shape specific parenting practices, Fagot (1978) and Leaper, Anderson, and Sanders (1998) reported the presence of variability in parenting practices based on children’s gender. They reported that mothers tend to praise and show more approval towards their daughters than towards their sons. In a large, representative survey of households in Northern Ireland, Lloyd and Devine (2006) found that girls under 17 years of age (especially the younger ones) received more positive parenting (expressed as hugging, cuddling, little yelling, and little ‘smacking’) while the boys were treated with more negative interactions (including more yelling). This trend in parenting is observed across cultures. In their study of gender differences and birth order on perceived parenting styles in Japan, Someya et al. (2000) reported strong birth order and gender effects on parenting styles. The eldest sons experienced the most rejecting parenting style than any of the other  166 combinations of birth order and gender, including first born daughters. There is some evidence that there is gender variation in parenting in the context of feeding-related practices, eating disorders and obesity (Birch et al., 2003 Hill & Franklin, 1998). For example, highly restrictive feeding practices have been associated with eating in the absence of hunger (Birch et al., 2003) and in overweight among girls (Birch & Fisher, 2000; Fisher & Birch, 2002). Gender differences concerning parental prompting and encouragement to eat have been reported among Anglo-American and Mexican-American families, with boys being encouraged to eat more than are girls (Klesges et al., 1986; Olvera- Ezzell et al., 1990). Klesges et al. (1983) reported significant correlations between children’s relative weight and parental encouragement and prompting and reported that parents contribute to the acceleration in weight. For example, Waxman and Stunkard (1980) observed the feeding practices of boys and their siblings and reported differential practices between obese and non-obese siblings, with mothers serving larger portions to their obese boys than to their normal weight boys. Conversely, overweight girls tend to elicit more restrictive feeding practices from their mothers than do non-overweight girls (Birch et al., 2003). The opposing effect of ineffective parenting practices leading to obesity in the boys and girls of the current study may be attributed to gender biases. Mothers are generally more concerned about their daughters’ weight than of their sons’ weight, especially if they themselves have problematic weight (Agras et al., 1999; Costanzo & Woody, 1984). The parenting practices of the PMKs who adopted negative disciplinary practices towards the girls, in this study, may be attributed to such concerns. However, causal relationships have not been explored and the literature is not clear about whether the observed parental concerns  167 started before the onset of obesity or whether the parenting practices emerged following the development of obesity. In conclusion, differential parenting for boys and girls suggest that overly critical and disapproving parenting practices have detrimental effects on girls’ weight status, and that parenting characterized as lax is also ineffective and problematic, particularly for boys. The studies reviewed are not clear about a causal relationship between parenting practices and the development of problematic weight. To better understand the emergence of obesity between boys and girls and its relationship with parenting practices, there is a need for future longitudinal studies that commence with the child’s birth weight and concurrently follow the parenting practices of the children’s parents (i.e., a developmental approach should be considered). Limitations and implications This last section deals with some of the methodological limitations of the study including the validity of the scales used and the validity of the reported anthropometric measures and some implications of the findings for nursing and future research. Methodological limitations A major finding of this study is the heterogeneous trajectories of BMI that would not have been revealed in either conventional cross-sectional studies or in growth models. This study explored unique variables that put children at risk for different trajectories of obesity and explicated some contextual differences experienced by obese children. However, caution must be taken when interpreting or generalizing the results because, as in most studies, several limitations exist. But most importantly, the results of this study apply only to the case  168 when temperament and obesity are first measured between the ages of 2-3 years and as predictors of later trajectories of obesity (up to 9 years of age). The following sections describe various limitations as they pertain to the validity of the measures used, the validity of parental reporting of anthropometric measures, and additional potential biases and confounders. The validity of the parenting practices and temperament scales  Although the statistical models, produced in the analyses presented here, were found to be significant, the magnitude of the effects was not impressive. A potential explanation for this outcome is that the NLSCY was not designed to answer questions such as those posed in this study and may not have provided optimum measures.62 There is some concern regarding the soundness of the measures, especially in light of the unsatisfactory internal consistency of the ineffective parenting measure (Cronbach’s alpha = 0.71) and the questionable factorial structure (dimensionality) of the temperament measure. Further studies are needed to replicate the findings of this study with more rigorous measures. Similar to the majority of studies in the field of socialization, the dimensions of parenting and the fussy/difficult temperament, in the NLSCY, were identified through exploratory factor analysis, which presents comparability and reliability challenges. First, making direct comparisons between contrasting theoretical positions (and study conclusions) is extremely difficult (i.e., the factorial structures of the scales used in the NLSCY could be  62 The scales used to measure parenting practices were developed specifically for the NLSCY and the psychometric properties of these scales are not well established. For example, there is little evidence that supports the construct validity of the scales or whether the items are good measurements of parental consistency, ineffective practices, or positive interactions. Furthermore, some of the questions were phrased with compound meanings, thus the responses of the participants were based on their subjective understanding of the questions, which could have affected the validity of the data.  169 different from the ones described in the literature) and second, the incapability of factor analysis to resolve theoretical differences, because of its subjective execution, is problematic. Further, a major limitation to this approach lies in the fact that the set of dimensions yielded by the analysis depended on the initial set of items included in the covariance matrix.63 This current study may have different conclusions if replicated with different measures or with different factor structures for the indicators. The 9-item ‘fussy/difficult’ temperament scale, which was developed specifically for this study, was developed through a combination of statistical analyses (exploratory factor analysis) and theoretical knowledge (based on Thomas and Chess’s (1977) description of emotionality). The nine items showed moderately strong factor loadings and moderate internal consistency (Cronbach’s alpha = 0.81). The reliability coefficient for the items generated from the factor analysis, in this study, which was labeled as the fussy/difficult dimension of temperament, was superior to that reported by Jabel et al. (2002) in their factorial analysis of the ‘good-natured’ temperament (Cronbach’s alpha = 0.78 for 8 items). Further assessment of reliability is nonetheless imperative. The temperament and parenting items were reported by questionnaire; some researchers aptly argue that there is a need for observational measures where context can be taken into account. Observational measures could shed light on the process of interaction between children and their parents. However, observational measures could not provide information about infrequent, albeit important, interactions (Maccoby, 2007). Therefore, a combination of the two approaches may provide a more balanced and complementary  63 For example, the rationale and process by which the 25 original parenting items were selected for inclusion is not explicated in the accompanying survey documentation. All that is understood of the decision is that the items were based on the work of Dr. Dodge and the adaptation of Strayhorn and Wiedman’s (1988) Parenting Practices Scale.  170 depiction of these interactions. The parents’ parenting practices and children’s temperament were measured simultaneously, and at one time, in this study. Although the literature provides evidence of the relative stability of these constructs, one-point assessments obviously cannot capture any developmental changes in the constructs. This may be desirable for the temperament construct where change is likely associated with personality development and not changes in temperament, per se, but parents do learn to be more effective parents over time and hence one-time assessments are limited. The validity of parental reports of children’s weight and height A major limitation to this study is the validity of the parents’ reports of their children’s heights and weights and the subsequent measurement of BMI. Errors in BMI estimation may have caused misclassification of the children into the various BMI trajectory groups and significantly under- or over-estimated the true prevalence of obesity (Niedhammer et al., 2000; Roberts, 1995). The ability of parents to accurately report their children’s height and weight is not well established. There have been few studies conducted to assess the accuracy of parental reporting and the results of the completed studies have been inconsistent. For example, Davis and Gergen (1994) and Huybrechts et al. (2006) reported that parental reports are inaccurate for the purpose of classifying preschool children into BMI categories while Sekine et al. (2002) concluded that parental reports are valid for the study of childhood obesity. Notwithstanding the published literature, the effects of any inaccurate reporting may have been reduced, to some extent, by the elimination of the biologically implausible outliers and by the fact that the children’s weight and height were reported every 2 years; repeated  171 measures may have encouraged the parents to keep records of their children’s weights and heights. The methods used in this study could also have reduced the impact of any inaccurate reporting by analyzing group means over time (i.e., group-based trajectories allowed for the estimation of change in BMI across time and were based on the groups’ mean BMIs, rather than the individuals’). Additional potential biases and confounders An important source of bias in this study resulted from the parenting practices and temperament items being rated by the same person (the PMK), which may therefore have reflected the respondents’ tendencies, contexts and expectations and led to confounded method variance. For example, depression and lack of social support are known to affect parental care and thus subsequent parental ratings. Parents with poor social support and stressful interpersonal relationships have been noted to have less sensitive and more intrusive styles of interaction with their preschool children (Jennings et al., 1990). Schaffer (1996) reviewed the literature on social support and its impact on parenting and concluded that “parents who receive more support behave towards their children with greater warmth and more consistency, are able to provide more effective discipline and yet be less punitive, respond to their children with great sensitivity, have more positive attitudes about child rearing, and show greater affection” (p. 231). Equally important, the effects of maternal depression can be seen relatively early in a child’s development; infants of depressed mothers are reported to be more fussy and irritable and have lower activity levels (Field, 1984), although it is not clear whether depression contributes to the fussiness or whether depression is triggered by fussy temperaments. However, one could argue that even in the presence of such bias, there is reason to believe  172 that the bias did not directly interfere with the findings and that the measures operated well because the variables were predictive of the girls’ obesity trajectory. What is likely is that there are other factors associated with the development of obesity that were not analyzed in this study. The findings of this study could have resulted from the confounding effects of other unmeasured variables associated with both temperament and parenting, such as feeding practices that have been shown to be influenced by temperament and parenting practices. Other potential unmeasured confounders include total caloric consumption, physical activity and sedentary behavior. More critically, no data about parental weight status was available and it is known that genetics play a large role in the determination of children’s weight status. Parents’ weight status also affects their perceptions of whether their children are obese (and may have influenced their reports of their children’s heights and weights). Agras et al. (2004) showed that children’s temperaments moderated the effects of parental weight status on the development of childhood obesity. Children with difficult temperaments and an overweight parent were at higher risk of developing obesity than were children with non- problematic temperaments and an overweight parent. Further, overweight parents of difficult children tended to use more negative and positive prompting about their children’s eating than did overweight parents with children with low risk temperaments. Implications The following section describes several potential implications for nursing practice and research arising from this study of childhood obesity. Because this is a new area of research, additional conceptual work is needed to identify the most important dimensions of temperament and parenting practices that are integral to the development of obesity;  173 identifying the relevant influences will require collaboration across diverse disciplinary teams. Implications for practice The results of this study present important implications for nursing practice. The findings support the need for nurses to be knowledgeable about the different ways in which obesity develops and the atypical but important risk factors contributing to childhood obesity. Understanding the different trajectories of BMI and their risk factors will allow nurses and other health professionals to take different measures and approaches in the assessment, intervention and evaluation of obesity and obesity-related health problems. For example, a child presenting with normal weight may be overlooked as being at risk for obesity and may not receive proper assessment and counseling. Yet early intervention may be the key to addressing this complex and somewhat intractable health problem. Community health nurses have a great opportunity to identify children at risk for obesity and to intervene at an early age. The findings of this study highlight the need for further development of nursing assessments and preventive approaches to obesity that include consideration of the biological and social contexts of children. Nursing interventions could focus on parenting programs that emphasize understanding children’s different temperaments and consequent parental reactions and optimal management. Interventions could also focus on enhancing parenting practices and behavior that optimize growth and development, which could extend to the larger social environment of children in schools and child-care facilities. If such interventions were found to be effective, in addition to avoiding the plight of obesity, children could develop positive social behavior that would allow them to deal with the vast array of life experiences that they will inevitably face.  174 Implications for future research: Measurement implications for parenting practices and temperament The association between parenting practices and childhood weight status is not clearly established, especially in terms of the role of ineffective parenting. In exploring the relationship between parenting practices and childhood obesity, much of the limited literature has relied upon different conceptualizations and measurements of parenting that have most likely contributed to the contradictory results that exist. In attempting to relate parenting practices to subsequent obesity, researchers have considered different types of parenting styles such as authoritarian, authoritative, democratic or permissive practices (Agras et al., 2004; Brewis, 2003; Chen, Liu, & Li, 2000; Rhee et al., 2006). A limited number of studies have explored the relationship through more extensive patterns of punitive parenting such as neglect and abuse (Lissau-Lund & Sфrensen, 1994; Mustillo et al., 2003). The limitations of using such broad categories to identify prevailing parenting styles are multiple but, most importantly, global styles of parenting lack information about the context or appropriateness under which such parenting occurs (Darling & Steinberg, 1993; Maccoby, 2007). Researchers have reported that the majority of parents utilize every one of the types of styles identified and their use is contingent on situational or contextual factors, such as their socialization goals and the child’s temperament (Grusec & Goodnow, 1994; Kochanska, 1990). Relevant to his study, Olvera-Ezzell, Power, and Cousins (1990) found that situational contexts may be more powerful determinants of socialization feeding strategies leading to obesity than are fixed parenting styles. In their study about the socialization of eating habits in Mexican-American children, they found that mothers use different strategies to control eating behavior based on their socialization goals; the mothers reported being permissive  175 when encouraging the child to eat new food, authoritarian when the child refused eat, and authoritative when discouraging eating. Future studies should be designed to examine specific feeding practices in addition to general parenting practices. Methodological implications for future research Currently, a major challenge in this field of study is the measurement of temperament and parenting practices as independent or related risk factors. Concerted efforts are needed to understand the contribution of both temperament and parenting practices as interacting agents in children’s growth. Longitudinal studies have shown that parenting practices are altered in response to children’s development (Chao & Willms, 2003; Collins & Madsen, 2003; Maccoby, 2007). In recent years, it has been suggested that the manifestations of temperament are not static but in a state of development (Bates & Pettit, 2007; Rothbart & Bates, 1998), although not much evidence is available to show how temperament may change or unfold over time (Halverson & Deal, 2001). A potentially new statistical tool that could capture the developmental aspects of parenting and temperamental change and how they influence one another is the use of dual trajectory analysis. According to Nagin (2005), the dual model is an extension of the group-based trajectory framework that moves from a single outcome to modeling the developmental course of two distinct but related outcomes. The dual trajectory provides a summary of the developmental association between two outcomes (for example temperament and parenting) by capturing the dynamic dimension of overlap between the outcomes of interest. The application of this model could capture the developmental evolvement of the variables of interest. The analysis is conducted by identifying distinctive clusters of developmental trajectories for temperament and parenting, separately, and then through joint trajectories; both conditional and joint probabilities of  176 trajectory membership across the two variables could be identified. Longitudinal research and additional contributing risk factors  Longitudinal research is needed to identify risk and protective factors for optimal childhood weight status. These designs should include samples that are ethnically and socio- economically diverse. Ideally, the research designs should include measures of physical and sedentary behavior and food intake and should assess aspects of the environment related to parents’ feeding strategies and the eating habits of the children. The designs could also expand on research related to temperament and parenting by incorporating measures of the psychological well being of the children (e.g., mental health, body concerns, eating disorders, emotional eating, depression and self esteem). Conclusion The seriousness of childhood obesity lies in its health, public, and economic consequences. Prevention of obesity is of utmost importance as intervention programs are showing marginal impact on weight management. To prevent obesity, there is a need to focus on identifying and understanding the risk factors that lead to problematic weight status. In addition to the documented risk factors associated with the high prevalence of childhood obesity (diet, exercise, and sedentary behavior), it also has been reported that childhood obesity is associated with both the family’s environment and the child’s intrinsic characteristics. For this study, the relationships between child temperament and specific parenting behaviors were explored to predict the development of overweight or obesity. The results of this study have the potential to further our understanding of factors associated with the development of obesity at a younger age and hence help in the development of early  177 preventive programs.     178 References  Agras, W. S., Kraemer, H. C., Berkowitz, R., Korner, A., & Hammer, L. D. (1987). Does a vigorous feeding style influence early development of obesity? Journal of Pediatrics, 110, 799-804. Agras, W. S., Hammer, L. D., & McNicholas, F. (1999). A prospective study of the influence of eating-disordered mothers on their children. International Journal of Eating Disorder, 25, 253-262. Agras, W. S., Hammer, L. D., McNicholas, F., & Kraemer, H. C. (2004). Risk factors for childhood overweight: A prospective study from birth to 9.5 years. Journal of Pediatrics, 145, 19-24. Anderson, S. E., Bandini, L. G., Dietz, W. H., & Must, A. (2004). Relationship between temperament, nonresting energy expenditure, body composition, and physical activity in girls. International Journal of Obesity, 28, 300-306. Anderson, S. E., Bandini, L. G., & Must, A. (2005). Child temperament does not predict adolescent body composition in girls. International Journal of Obesity, 29, 47-53. Arredondo, E., Elder, J., Ayala, G., Campbell, N., Baquero, B., & Duerksen, S. (2006). Is parenting style related to children’s healthy eating and physical activity in Latino families? Health Education Research, 21, 862-871. Austin, M., Hadzi-Pavlovi, D., Leader, L., Saint, K., & Gordon, P. (2005). Maternal trait anxiety, depression and life event stress in pregnancy: Relationships with infant temperament. Early Human Development, 81, 183-188. Baranowski, T., Bouchard, C., Bar-Or, O., Bricker, T., Heath, G., Kimm, S. Y., et al. (1992).  179 Assessment, prevalence, and cardiovascular benefits of physical activity and fitness in youth. Medicine & Science in Sports & Exercise, 24(Suppl. 6), 237-247. Barber, B. K. (1996). Parental psychological control: Revisiting a neglected construct. Child Development, 67, 3296-3319. Barber, B. K. (1997). Introduction: Adolescent socialization in context-connection, regulation, and autonomy in multiple contexts. Journal of Adolescent Research, 12, 173- 177. Barber, B. K., & Olson, J. A. (1997). Socialization in context: Connection, regulation, and autonomy in the family, school, neighborhood, and with peers. Journal of Adolescent Research, 12, 287-315. Barlow, S. E., & Dietz, W. H. (1998). Obesity evaluation and treatment: Expert committee recommendations. The Maternal and Child Health Bureau, Health Resources and Services Administration and the Department of Health and Human Services [Electronic version]. Pediatrics, 102(3), E29. Bates, J. E. (1980). The concept of difficult temperament. Merrill-Palmer Quarterly, 26, 299-319. Bates, J. E. (1989). Concepts and measures of temperament. In G. A. Kohnstamm, J. E. Bates, & M. K. Rothbart (Eds.), Temperament in childhood (pp. 3-33). Chichester, UK: John Wiley. Bates, J. E., & Bayles, K. (1984). Objective and subjective components in mothers’ perceptions of their children from age 6 months to 3 years. Merrill-Palmer Quarterly, 30, 111-130. Bates, J. E., Bennett-Freeland, C. A., & Lounsbury, M. L. (1979). Measurement of infant  180 difficulties. Child Development, 50, 794-803. Bates, J. E., Dodge, K. A., Pettit, G. S., & Ridge, B. (1998). Interaction of temperamental resistance to control and restrictive parenting in the development of externalizing behavior. Developmental Psychology, 34, 982-995. Bates, J. E., Olson, S., Pettit, G., & Bayles, K. (1982). Dimensions of individuality in the mother-infant relationships at six months of age. Child Development, 53, 446-461. Bates, J. E., & Pettit, G. S. (2007). Temperament, parenting, and socialization. In J. E.  Grusec & P. D. Hastings (Eds.). Handbook of socialization: Theory and research (pp. 153-177). New York: The Guilford Press. Baughcum, A. E., Burklow, K., Deeks, C., Powers, S. W., & Whitaker, R. (1998).  Maternal feeding practices and childhood obesity. Archives of Pediatrics and Adolescent Medicine, 152, 1010-1014. Baughcum, A. E., Powers, S. W., Johnson, S., Khamberlin, L., Deeks, C., Joran, A., et al. (2001). Maternal feeding practices and beliefs and their relationships to overweight in early childhood. Developmental and Behavioral Pediatrics, 22, 391-408. Baumrind, D. (1967). Child care practices anteceding three patterns of preschool behavior. Genetic Psychology Monographs, 75, 43-88. Baumrind, D. (1991). Parenting styles and adolescent development. In J. Brooks-Gunn, R. Lerner, & A. C. Peterson (Eds.), The encyclopedia of adolescence (pp. 746-758). New York: Garland. Baumrind, D. (1996). The discipline controversy revisited. Family Relations, 45, 405- 414. Bavdekar, A., Yajnik, C. S., Fall, C. H. D., Bapat, S., Pandit, A. N., Deshpande, V., et al.  181 (1999). Insulin resistance in 8-year old Indian children: Small at birth, big at 8 years, or both? Diabetes, 48, 2422-2429. Bellizzi, M. L., & Dietz, W. H. (1999). Workshop on childhood obesity: Summary of the discussion. American Journal of Clinical Nutrition, 70(Suppl.), 173-175. Belsky, J., Hsieh, K., & Crnic, K. (1998). Mothering, fathering and infant negativity as antecedents of boys’ externalizing problems and inhibition at age 3 years: Differential susceptibility to rearing experience? Development and Psychopathology, 10, 301-319. Berenson, G. S., Srinivasan, S. R., Bao, W., Newman, W. P., Tracy, R. E., & Wattigney, W. A. (1998). Association between multiple cardiovascular risk factors and atherosclerosis in children and young adults. New England Journal of Medicine, 338, 1650-1656. Bernard-Bonnin, A. C., Gilbert, S., Rousseau, E., Masson, P., & Matheux, B. (1991). Television and the 3- to 10-year old child. Pediatrics, 88, 48-54. Betz, C. (2000). Childhood obesity: Nursing prevention and intervention approaches are needed. Journal of Pediatric Nursing, 15, 135-136. Biddle, S. J., Gorely, T., & Stensel, D. J. (2004). Health-enhancing physical activity and  sedentary behaviour in children and adolescents. Journal of Sports Sciences, 22, 679- 701. Birch, L. L. (1998). Psychological influences on the childhood diet. Journal of Nutrition, 12(Suppl. 2), 407-410. Birch, L. L., & Davison, K. K. (2001). Family environmental factors influencing the developing behavioral controls of food intake and childhood overweight. Pediatric Clinics of North America, 48, 893-907.  182 Birch, L. L., & Fisher, J. O. (1995). Appetite and eating behavior in children. Pediatrics Clinics in North America, 42, 931-952. Birch, L. L., & Fisher, J. O. (1996). The role of experience in the development of  children’s eating behavior. In E. D. Capaldi (Ed.), Why we eat what we eat: The psychology of eating (pp. 113-141). Washington, DC: American Psychological Association. Birch, L. L., & Fisher, J. O. (1998). Development of eating behaviors among children and adolescents. Pediatrics, 101, 539-549. Birch, L. L., & Fisher, J. O. (2000). Mothers’ child-feeding practices influence daughters’ eating and weight. American Journal of Clinical Nutrition, 71, 1054-1061. Birch, L. L., Fisher, J. O., & Davison, K. K. (2003). Learning to overeat: Maternal use of restrictive feeding practices promotes girls’ eating in the absence of hunger. American Journal of Clinical Nutrition, 78, 215-220. Birmingham, C. L., Muller, J. L., Palepu, A., Spinelli, J. J., & Anis, A. L. (1999). The cost of obesity in Canada. Canadian Medical Association Journal,160, 483-488. Borkowski, J. G., Ramey, S., & Bristol-Power, M. (2002). Parenting and the child’s world: Influences on academic, intellectual, and social-emotional development. Mahwah, NJ: Erlbaum. Bornstein, M. H. (1995a). Parenting infants. In M. H. Bornstein (Ed.), Handbook of  parenting: Vol 1. Children and parenting (pp. 3-40). New Jersey: Lawrence Erlbaum. Bornstein, M. H. (1995b). Preface. In M. H. Bornstein (Ed.), Handbook of parenting: Vol. 2.  Biology and ecology of parenting (pp. xiii-xv). New Jersey: Lawrence Erlbaum. Brand, H., Crous, B., & Hanekom, J. (1990). Perceived parental inconsistency as a  183  factor in the emotional development of behaviour-disordered children. Psychological Reports, 66, 620-622. Brewis, A. (2003). Bicultural aspects of obesity in young Mexican schoolchildren. American Journal of Human Biology, 15, 446-460. Buiten, C., & Metzger, B. (2000). Childhood obesity and risk of cardiovascular disease: A review of the science. Pediatric Nursing, 26(1), 13-18. Bundred, P., Kitchiner, D., & Buchan, I. (2001). Prevalence of overweight and obese children between 1989 and 1998: Population based series of cross-sectional studies. BMJ, 322, 326-328. Buss, A. H., & Plomin, R. (1984). Temperament: Early developing personality traits. Hillsdale, NJ: Erlbaum. Bussey, B., & Morgan, S. (1997). Obesity: Is there effective treatment now? Consultant, 37, 2945-2957. Cairns, R. B., & Cairns, D. D. (1991). Social cognition and social networks: A developmental perspective. In D. J. Pepler & K. H. Rubin (Eds.), The development and treatment of childhood aggression (pp. 249-278). Hillsdale, NJ: Erlbaum. Cameron, J. R. (1977). Parental treatment, children’s temperament and the risk of childhood behavioral problems: Relationships between parental characteristics and changes in children’s temperament over time. American Journal of Orthopsychiatry, 47, 568-576. Campbell, S. (1979). Mother-infant interaction as a function of maternal ratings of temperament. Child Psychiatry and Human Development, 10, 67-76. Campbell, K. J., Waters, E., O’Meara, S., Kelly, S., &. Summerbell, C. D. (2004).  184 Interventions for preventing obesity in children. The Cochrane Library, 3, CD00187. Canning, P., Courage, M., & Frizzell, L. (2004). Prevalence of overweight and obesity in a provincial population of Canadian preschool children. CMAJ, 171, 240-242. Carey, W. B. (1985). Temperament and increased weight gain in infants. Developmental and Behavioral Pediatrics, 6, 128-131. Carey, W. B., Hegvik, C., & McDevitt, S. C. (1988). Temperament factors associated with rapid weight gain and obesity in middle childhood. Developmental and Behavioral Pediatrics, 9, 194-198. Casey, P., Szeto, K., Lensing, S., Bogle, M., & Weber, J. (2001). Children in food- insufficient, low-income families. Archives of Pediatrics and Adolescent Medicine, 155, 508-514. Centers for Disease Control and Prevention. (2005). A SAS program for the CDC growth charts. Retrieved 2006, from: http://cdc.gov/nccdphp/dnpa/growthcharts/sas.htm. Chao, R., & Willms, D. (2003). The effects of parenting practices on children’s outcomes. In D. Willms (Ed.), Vulnerable children: Findings from Canada’s National Longitudinal Survey of Children & Youth (pp. 149-165). Edmonton, AB: University of Alberta Press. Chen, J-L., & Kennedy, C. (2004). Family functioning, parenting style and Chinese children’s weight status. Journal of Family Nursing, 10, 262-279. Chen, X., Liu, M., & Li, D. (2000). Parental warmth, control, and indulgence and their  relations to adjustment in Chinese children: A longitudinal study. Journal of Family Psychology, 14, 401-419. Chess, S., & Thomas, A. (1984). Origins and evolution of behavior disorders. New York: Brunner-Mazel.  185 Chess, S., & Thomas, A. (1989). The practical application of temperament to psychiatry. In W. B. Carey & S. C. McDevitt (Eds.), Clinical and educational applications of temperament research (pp. 23-35). Berwyn, PA: Swets North America. Chess, S., & Thomas, A. (1996). Temperament: Theory and practice. New York: Brunner/Mazel. Chess, S., & Thomas, A. (1999). The development of behavioral individuality. In M. D. Levine, W. C. Carey, & A. C. Crocker (Eds.), Developmental-behavioral pediatrics (pp. 89-99). Philadelphia: W. B. Saunders. Christoffel, K. K., & Forsyth, B. W. (1989). Mirror image of environmental deprivation: Severe childhood obesity of psychological origin. Child Abuse and Neglect, 13, 249-256. Cole, T. J., Bellizzi, M. C., Flegal, K. M., & Dietz, W. H. (2000). Establishing a standard definition for child overweight and obesity worldwide: International survey. BMJ, 320, 1240-1243. Collins, W. A., & Madsen, S. (2003). Developmental change in parenting interactions. In L. Kucynski (Ed.), Handbook of dynamics in parent-child relations (pp. 49-66). Thousand Oaks, CA: Sage. Cooper, D. M., Poage, J., Barstow, T., & Springer, C. (1990). Are obese children truly unfit? Minimizing the confounding effect of body size on exercise response. Journal of Pediatrics, 116, 223-230. Costa, P., & McCrae, R. R. (2001). A theoretical context for adult temperament. In T.  Wachs, & G. Kohnstamm (Eds.), Temperament in context (pp.1-21). New Jersey: Lawrence Erlbaum.  186 Costanzo, P. R., & Woody, E. Z. (1984). Parental perspectives on obesity in children: The importance of sex difference. Journal of Social and Clinical Psychology, 2, 305-313. Costello, E. J., Keeler, G. P., & Angold, A. (2001). Poverty, race/ethnicity, and psychiatric disorder: A study of rural children. American Journal of Public Health, 91, 1494-1498. Cragg, S., Cameron, C., Craig, C. L., & Russell, S. (1999). Canada’s children and youth: A physical activity profile. Special Report Series. Canadian Fitness and Lifestyle Research Institute. Retrieved March, 2006, from www.cflri.ca/pdf/e/98NLSCY.pdf Crockenberg, S., & Smith. (1982). Antecedents of mother-infant interaction and infant irritability in the first three months of life. Infant Behavior Development, 5, 105-119. Dahinten, V. S., Shapka, J. & Willms, J. D. (2007). Adolescent children of adolescent  mothers: The impact of family functioning on trajectories of development. Journal of  Youth and Adolescence, 36, 195-212. Dahinten, V. S., & Willms, J. D. (2002). The effects of adolescent child-bearing on children’s outcomes. In J. D. Willms (Ed.), Vulnerable children: Findings from Canada’s National Longitudinal Survey of Children & Youth (pp. 243-258). Edmonton, AB: University of Alberta Press. Danielzik, S., Czerinski-Mast, M., Langnase, K., Dilba, B., & Muller, M.J. (2004). Parental overweight, socioeconomic status, and high birth weight are the major determinants of overweight and obesity in 5-7 year old children: Baseline data of the Kiel obesity prevention study (KOPS). International Journal of Obesity, 28, 1494-1502. Darling, N., & Steinberg, L. (1993). Parenting style as context: An integrative model. Psychological Bulletin, 113, 487-496.  187 Davies, P. S., Day, J. M., & Lucas, A. (1991). Energy expenditure in early infancy and later body fatness. International Journal of Obesity, 15, 727-731. Davies, P. S., Gregory, J., & White, A. (1995). Energy expenditure in children aged 1.5-4.5 years: A comparison with current recommendations for intake. European Journal of Clinical Nutrition, 49, 360-364. Davis, H., & Gergen, P. J. (1994). Mexican-American mothers’ reports of the weights and heights of children 6 months through 11 years old. Journal of the American Dietetic Association, 94, 512-16. Davison, K. K., & Birch, L. L. (2001). Child and parent characteristics as predictors of change in girls’ body mass index. International Journal of Obesity, 25, 1834-1842. De Bourdeaudhuij, I., & Van Oost, P. (2000). Personal and family determinants of dietary behavior in adolescents and their parents. Psychology and Research, 15, 751-770. Decaluwe, V., Braet, C., Moens, E., & Vlierberghe, L. (2006). The association of  parental characteristics and psychological problems in obese youngsters. International   Journal of Obesity, 30, 1766-1774. DeFronzo, R. A., & Ferrannini, E. (1991). Insulin resistance: A multifaceted syndrome responsible for NIDDM, obesity, hypertension, dyslipidemia, and atherosclerotic vascular disease. Diabetes Care, 14, 173-194. Deheeger, M., Rolland-Cachera, M. F., & Fontvielle, A. M. (1997). Physical activity and body composition in 10-year old French children: Linkages with nutritional intake? International Journal of Obesity and Related Metabolic Disorders, 21, 372-379. De Vries, M. W. (1984). Temperament and infant mortality among the Masai of East Africa. American Journal of Psychiatry, 141, 1189-1194.  188 Dietitians of Canada, Canadian Pediatric Society, Community College of Family Physicians of Canada, and the Community Health Nurses Association of Canada. (2004). The use of growth charts for assessing and monitoring growth in Canadian infants and children. Canadian Journal of Dietetic Practice and Research, 65(1), 22-32. Dietz, W. (1998). Health consequences of obesity in youth: Childhood predictors of adult disease. Pediatrics, 101, 518-525. Dietz, W. (2000). “Adiposity rebound”: Reality or epiphenomenon? Lancet, 356(9247), 2027-2028. Dietz, W., & Gortmaker, S. (2001). Preventing obesity in children and adolescents. Annual Review of Public Health, 22, 337-353. Dietz, W. H., & Robinson, T. N. (1998). The use of the body mass index as a measure of overweight in children and adolescents. Journal of Pediatrics, 132, 191-193. Diez-Roux, A. V. (1998). Bringing context back into epidemiology: Variables and fallacies in multilevel analysis. American Journal of Public Health, 88, 216-222. Eisenberg, N., Guthrie, I. K., Fabes, R. A., Sheppard, A., Losoya, S., Murphy, B. C., et al. (2000). Prediction of elementary school children’s externalizing problem behaviors from attentional and behavioral regulation and negative emotionality. Child Development, 71, 1367-1382. Eriksson, J. G., Forsen, C., Osmond, T. J., & Barker, D. J. (2003). Pathways of infant and childhood growth that lead to type 2 diabetes. Diabetes Care, 26, 3006-3010. Evans, R. G., Barer, M. I., & Marmor, T. R. (Eds.). (1994). Why are some people healthy and others not? The determinants of health of populations. New York: Walter de Gruyter.  189 Fagot, B. I. (1978). The influence of sex of child on parental reactions to toddler children. Child Development, 49, 459-465. Faith, M., Berkowitz, R., Stallings, V., Kerns, J., Storey, M., & Stunkard, A. (2004). Parental feeding attitudes and styles and child body mass index: Prospective analysis of a gene-environment interaction. Pediatrics, 114, 429-436. Feldman, W., Feldman, E., & Goodman, J. (1988). Culture versus biology: Children’s attitudes towards thinness and fatness. Pediatrics, 81, 190-194. Felitti, V. J., Anda, R. F., Nordenberg, D., Williamson, D. F., Spitz, A. M., Edwards, V., et al. (1998). Relationship of child abuse and household dysfunction to many of the leading causes of death in adults. American Journal of Preventive Medicine, 14, 245-258. Ferguson, M. A., Gutin, B., Le, N., Karp, W., Humphries, M., Okuyama, T., et al. (1999). Effects of exercise training and its cessation on components of the insulin resistance syndrome in obese children. International Journal of Obesity, 22, 889-895. Fergusson, D.M., & Woodward, L. (1999). Maternal age and educational and psychosocial outcomes in early adulthood. Journal of Child Psychology and Psychiatry, 43, 479-489. Field, T. M. (1984). Early interactions between infants and their postpartum depressed mother. Infant Behavior and Development, 7, 517-522. Field, A. E., Camargo, C. A., Taylor, C. B., Berkey, C. S., Roberts, S. B., & Colditz, G. A. (2001). Peers, parents, and media influences on the development of weight concerns and frequent dieting among pre-adolescents and adolescent girls and boys. Pediatrics, 107, 54-60. Finegan, J. A., Niccols, A., Zacher, J., & Hood, J. (1989). Factor structure of the preschool characteristics questionnaire. Infant Behavior and Development, 12, 221-227.  190 Fish, M., & Crockenberg, S. (1981). Correlates and antecedents of nine-month infant behavior and mother-infant interaction. Infant Behavior and Development, 4, 69-81. Fisher, J. O., & Birch, L. (1999). Restricting access to palatable foods affects children’s behavioral response, food selection and intake. American Journal of Clinical Nutrition, 69, 1264-1272. Fisher, J. O., & Birch, L. (2000). Parents’ restrictive feeding practices are associated with  young girls’ negative self-evaluation of eating. Journal of the American Dietetic Association, 100, 1341-1346. Fisher, J. O., & Birch, L. (2002). Eating in the absence of hunger and overweight in girls at 5 and 7 years of age. American Journal of Clinical Nutrition, 76, 226-231. Fisher, J. O., Mitchell, D. C., Smiciklas-Wright, H., & Birch, L. L. (2002). Parental influences on young girls’ fruit and vegetable, micronutrient, and fat intakes. Journal of the American Dietetic Association, 102, 58-64. Flegal, K. M., Ogden, C. L., Wei, R., Kuczmarski, R., & Johnson, C. (2001). Prevalence of overweight in US children: Comparison of US growth charts from the Centers for Disease Control and Prevention with other reference values for body mass index. American Journal of Clinical Nutrition, 73, 1086-1093. Fontvieille, A. M., Harper, I. T., Ferraro, R. T., Spraul, M., & Ravussin, E. (1993). Daily energy expenditure by five-year-old children, measured by doubly-labeled water. Journal of Pediatrics, 123, 200-207. Freedman, D., Dietz, W., Srinivasan, S., & Berenson, G. (1999). The relation of overweight to cardiovascular risk factors among children and adolescents: The Bogalusa Heart Study. Pediatrics, 103, 1175-1182.  191 Frick, P., Christian, R., & Wooton, J. (1999). Age trends in the assessment between parenting practices and conduct problems. Behavior Modification, 23, 106-128. Friedlander, S., Larkin, E., Rosen, C., Palermo, T., & Redline, S. (2003). Decreased quality of life associated with obesity in school age children. Archives of Pediatrics and Adolescent Medicine, 157, 1206-1211. Friedman, M. A., & Brownell, K. D. (1995). Psychological correlates of obesity: Moving to the next generation. Psychological Bulletin, 117, 3-20. Froidevoux, F., Schutz, Y., Christine, L., & Jequier, E. (1993). Energy expenditure in obese women before and during weight loss, after refeeding, and in the weight-relapse period. American Journal of clinical Nutrition, 57, 35-42. Gable, S., & Lutz, S. (2000). Household, parent, and child contributions to childhood obesity. Family Relations, 49, 293-300. Gallagher, D., Visser, M., Sepulveda, D., Pierson, R. N., Harris, T., & Heymsfield, S. B. (1996). Body mass index as an estimate of fatness across gender, age, sex, and ethnic groups. American Journal of Epidemiology, 143, 228-239. Garn, S. M., & Clark, D. (1975). Nutrition, growth, development, and maturation: Findings of the Ten State Nutritionally Survey of 1968-1970. Pediatrics, 56, 306-319. Gauthier, B. M., Hichner, J., & Ornetein, S. (2000). High prevalence of overweight  children and adolescents in the Practice Partner Research Network. Archives of Pediatrics and Adolescent Medicine, 154, 625-628. Gazzangia, J. M., & Burn, T. L. (1993). Relationship between diet composition and body fatness, with adjustment for resting energy expenditure and physical activity in preadolescent children. American Journal of Clinical Nutrition, 58, 21-28.  192 Gentile, D., & Walsh, D. (2002). A normative study of family media habits. Journal of  Applied Developmental Psychology, 23, 157-178. Gill, T. D. (1997). Key issues in the prevention of obesity. British Medical Bulletin, 53, 359- 388. Gilliom, M., Shaw, D., Beck, J., Schonberg, M., & Lukon, J. (2002). Anger regulation in dis- advantaged preschool boys: Strategies, antecedents, and the development of self control. Developmental Psychology, 38, 222-235. Gillman, M. W., Rifas-Shiman, S. L., Camargo, C. A., Berkey, C. S., Frazier, A. L., Rockett, H. R., et al. (2001). Risk of overweight among adolescents who were breastfed as infants. JAMA, 285, 2461-2467. Goldsmith, H. H., & Campos, J. J. (1986). Fundamental issues in the study of early temperament: The Denver twin temperament study. In M. E. Lamb, A. L. Brown, & B. Rogoff (Eds.), Advances in developmental psychology: Vol. 4. (pp. 231-283). Hillsdale, NJ: Lawrence Erlbaum. Gordon, B. N. (1981). Child temperament and adult behavior: An exploration of “goodness of fit.” Child Psychiatry and Human Development, 11, 167-178. Gortmaker, S. L., Must, A., Perrin, J. M., Sobol, A. M., & Dietz, W. H. (1993). Social and economic consequences of overweight in adolescents and young adulthood. New England Journal of Medicine, 329, 1008-1012. Grimm, L., & Yarnold, P. (1995). Introduction to multivariate statistics. In Reading and understanding multivariate statistics (pp. 1-18). Washington, DC: American Psychological Association. Grolnick, W. S. (2003). The psychology of parental control: How well-meant parenting  193 backfires. New Jersey: Erlbaum. Grummer-Strawn, L. M., Garza, C., & Johnson, C. L. (2002). Childhood growth charts. Pediatrics, 109, 141-142. Grusec, D., & Goodnow, J. (1994). Impact of parental discipline on the child’s internalization of values: A re-conceptualization of current points of view. Developmental Psychology, 30, 4-13. Guerin, D., Gottfried, A., & Thomas, G. W. (1997). Difficult temperament and behavioral problems: A longitudinal study from 1.5 to 112 years. International Journal of Behavioral Development, 21, 71-90. Guillaume, M., Lapidus, L., Becker, F., Lambert, A., & Bjorntorp, P. (1995). Familial trends of obesity through three generations: The Belgian-Luxemburg child study. International Journal of Obesity Related Metabolic Disorder, 19(Suppl.), 5-9. Gunnarsdottir, I., & Thorsdottir, I. (2003). Relationship between growth and feeding in infancy and body mass index at the age 6 years. International Journal of Obesity, 27, 1523-1527. Guo, S., & Chumlea, W., (1999). Tracking of body mass index in children in relation to overweight in adulthood. American Journal of Clinical Nutrition, 70(Suppl.), 145-148. Guo, S. S., Riche, A. F., Chumlea, W.C., Gardner, J. D., & Siervogel, R. M. (1994).   The predictive value of childhood body mass index for overweight at 35 years. American Journal of Clinical Nutrition, 59, 810-819. Halverson, C., & Deal, E. (2001). Temperamental change, parenting and the family context. In T. Wachs & G. Kohnstamm (Eds.), Temperament in context (pp. 61-79). New Jersey: Lawrence Erlbaum.  194 Harris, M., & Smith, S. (1983). The relationships of age, sex, ethnicity, and weight to stereotypes of obesity and self- perception. International Journal of Obesity, 7, 361-371. He, Q., Ding, Z., Fong, D., & Karlberg, J. (2000). Risk factors of obesity in preschool children in China: A population-based case-control study. International Journal of Obesity, 24, 1528-1536. Heo, M., Faith, M. S., Mott, J. W., Gorman, B. S., Redden, D. T., & Allison, D. B. (2003). Tutorial in biostatistics: Hierarchical linear models for the development of growth curves. An example with body mass index in overweight/obese adults. Statistics in Medicine, 22, 1911-1942. Hill, A. J., & Franklin, J. A. (1998). Mothers, daughters, and dieting: Investigating the transmission of weight control. British Journal of clinical Psychology, 37, 3-13. Hoffman, M. L. (1970). Moral development. In P. H. Mussen (Ed.), Carmichael’s manual of child psychology (Vol. 2, pp. 261-359). New York: Wiley & Sons. Humphrey, L. L. (1989). Observed family interactions among subtypes of eating disorders using structural analysis of social behaviour. Journal of Consulting and Clinical Psychology, 57, 206-214. Huston, A. C. (1983). Sex-typing. In E. M. Hetherington & P. H. Mussen (Eds.),  Handbook of child psychology: Vol. 4. Socialization, personality, and social development (pp. 387-467). New York: Wiley. Huybrechts, I., de Bacquer, D., Trimpont, I., De Backeer, G., & de Henauw, S. (2006).  Validity of parentally reported weight and height for preschool-aged children in Belgium and its impact on classification into body mass index categories. Pediatrics, 118, 2109- 2118.  195 Jabel, C., Normand, C., Tremblay, R., & Willms, D. (2002). Identifying vulnerable children at an early age. In J. D. Willms (Ed.), Vulnerable children: Findings from Canada’s National Longitudinal Survey of Children & Youth (pp. 105-120). Edmonton, AB: University of Alberta Press. Jain, A., Sherman, S. N., Chamberlin, L. A., Carter, Y., Powers, R., & Whitaker, C. (2001). Why don’t low-income mothers worry about their preschool children being overweight? Pediatrics, 107, 1138-1146. Janssen, I., Craig, W., Boyce, W. F., & Pickett, W. (2004). Associations between overweight and obesity with bullying behaviors in school-aged children. Pediatrics, 113, 1187-1194. Jennings, K., Stagg, V., & Connors, R. (1991). Social networks and mothers’ interactions with their preschool children. Child Development, 62, 966-978. Johnson, S. L., & Birch, L. L. (1994). Parents’ and children’s adiposity and eating style. Pediatrics, 94, 653-660. Johnson, S. L., Birch, L. L., & McPhee, L. (1991). Conditioned preferences: Young children prefer flavors associated with high dietary fat. Physiology and Behavior, 50, 1245-1251. Johnson-Down, L., O’Loughlin, K., & Gray-Donald, K. (1997). High prevalence of obesity in low income and multiethnic schoolchildren: A diet and physical activity assessment. Journal of Nutrition, 127, 2310-2315. Jones, B. (2005). PROC Traj. Retrieved May, 2005, from  http://www.andrew.cmu.edu/user/bjones/index.htm Jones, B., Nagin, D., & Roeder, K. (2001). A SAS procedure based on mixture models  196 for estimating developmental trajectories. Sociological Research and Methods, 29, 374- 393. Jonides, L., Buschbacher, V., & Barlow, S. (2002). Management of child and adolescent obesity: Psychological, emotional, and behavioral assessment. Pediatrics, 110, 215-221. Karam, J. H. (1992). Type II diabetes and syndrome X. Pathogenesis and glycemic management. Endocrinology and Metabolism Clinics of North America, 21, 329-350. Kass, R., & Wasserman, L. (1995). A reference Bayesian test for nested hypotheses and its relationship to Shwarz Criterion. Journal of American Statistical Association, 90, 928- 934. Katzmarzky, P. T., & Janssen, I. (2004). The economic costs associated with physical inactivity and obesity in Canada: An update. Canadian Journal of Applied Physiology, 29, 90-115. Katzmarzky, P. T., Persusse, L., Malina, R. M., Bergeron, J., Despres, J., & Bouchard, C. (2001). Stability of indicators of the metabolic syndrome from childhood and adolescence to young adulthood: The Quebec family study. Journal of Clinical Epidemiology, 54, 190-195. Keltikangas-Järvinen, L., Räikkönen, K., & Solakivi, T. (1990). Type A factors as predictors of somatic risk factors of coronary heart disease in young Finns: A six-year follow up study. Journal of Psychosomatic Research, 34, 89-97. Kensara, O. A., Wootton, S. W., Phillips, D. I., Patel, M., & Elia, M. (2004). Does body mass index reflect percentage body fat and body fat distribution in low and high weight subjects? Asia Pacific Journal of Clinical Nutrition, 13(Suppl.), 99. Kiess, W., Müller, G., Kapellen, T., & Böttner, A. (2001). Obesity in childhood and  197 adolescence: Epidemiology, etiology, clinical diagnosis and management. International Seminars in Pediatric Gastroentology and Nutrition, 10(4), 3-8. Kinston, W., Loader, P., & Miller, L. (1987). Emotional health of families and their members where a child is obese. Journal of Psychosomatic Research, 31, 583-599. Kinston, W., Loader, P., Miller, L., & Rein, L. (1988). Interaction in families with obese children. Journal of Psychosomatic Research, 32, 513-532. Kimm, S., Obarzanek, E., & Barton, B. (1996). Race, socioeconomic status, and obesity  in 9 to 10 year  old girls: The NHLB growth and health study. Annals of Epidemiology, 6, 266-275. Klesges, R. C., Coates, T. J., Brown, G., Sturgeonn-Tillisch, J., Moldenhauer-Klesges, L., Holtzer, B., et al. (1983). Parental influences on children’s eating behavior and relative weight. Journal of Applied Behavior Analysis, 16, 371-378. Klesges, R. C., & Hanson, C. L. Determining environmental causes and correlates (1988). In N. A. Krasnegor, G. D. Grave, & N. Kretchmer (Eds.), Childhood obesity: A biobehavioral perspective (pp. 89-118). New Jersey: Telford Press. Klesges, R. C, Klesges, L. M., Eck, L. H., & Shelton, M.L. (1995). A longitudinal analysis of accelerated weight gain in preschool children. Pediatrics, 95, 126-130. Klesges, R. C., Marlott, J. M., Boschee, P. F., & Weber, J. M. (1986). The effect of parental influences on children’s food intake, physical activity, and relative weight. International Journal of eating Disorders, 2, 335-346. Kochanska, G. (1990). Beliefs as long term predictors of mother-child interaction and report. Child development, 61, 1934-1940. Kohnstamm (1989). Temperament in childhood. In G. Bates, & M. K. Rothbart (Eds.),  198 Temperament in childhood (pp. 483-508). New York: John Wiley & Sons. Kotani, K., Nishida, M., Yamashita, S., Funstshi, T., Fujioka, S., Tokunaga, K., et al. (1997). Two decades of annual medical examinations in Japanese obese children: Do obese children become obese adults? International Journal of Obesity, 21, 912-921. Kramer, M. Barr, R., Pless, I., Bolisjoyly, C., McVey-White, L., & Leduc, D. (1986). Determinants of weight and adiposity in early childhood. Canadian Journal of Public Health, 77, 98-103. Kremers, S. P. J., Bruga, J., Hein de Vriesa, D., & Rutger, C. M. (2003). Parenting style and adolescent fruit consumption. Appetite, 41, 43-50. Kromeyer-Hauschild, K., Zeller, K., Jaeger, U., & Hoger, H. (1999). Prevalence of overweight and obesity among school children in Jena (Germany). International Journal of Obesity, 23, 1143-1150. Kuczmarski, R. J., Ogden, L. M., & Grummer-Strawn, K. M. (2000). CDC growth charts: United States. Advance Data from Vital and Health Statistics, 314, 1-28. Lacourse, E., Nagin, D., Tremblay, R. E., Vitaro, F., & Claes, M. (2003). Developmental trajectories of boys’ delinquent membership and facilitation of violent behaviors during adolescence. Development & Psychology, 15, 183-197. Laitinen, J., Power, C., Sovio, E., & Jarvelin, M. R. (2002). Unemployment and obesity among young adults in a northern Finland 1966 birth cohort. International Journal of Obesity, 26, 1329-1338. Leaper, C., Anderson, K. J., & Sanders, P. (1998). Moderators of gender effects on parents’ talk to their children: A meta analysis. Developmental Psychology, 34, 3-27. Ledingham, J. E. (1991). Social cognition and aggression. In D. J. Pepler & K. H. Rubin  199 (Eds.), The development and treatment of childhood aggression (pp. 279-286). Hillsdale, NJ: Erlbaum. Lee, C. L., & Bates, J. E. (1985). Mother-child interaction at age two years and perceived difficult temperament. Child Development, 56, 1314-1325. Lees, N. B., & Tinsley, B. J. (1998). Patterns of parental socialization of the preventive health behavior of young Mexican origin children. Journal of Applied Development Psychology, 19, 503-525. Lerner, R. M., Castellino, D. R., Terry, P. A., Villarruel, F. A., & McKinney, M .H. (1995). Developmental contextual perspective on parenting. In M. H. Bornstein (Ed.), Handbook of parenting: Vol. 2. Biology and ecology of parenting (pp. 285-309). Mahwah, NJ: Erlbaum. Li, R., O’Conner, L., Buckley, D., & Specker, B. (1995). Relation of activity level to body fat in infants 6 to 12 months of age. Journal of Pediatrics, 126, 353-357. Lissau-Lund, I., & Sфrensen, T. I. (1992). Prospective study of the influence of social factors in childhood on the risk of overweight in young adulthood. International Journal of Obesity, 16, 169-175. Lissau-Lund, I., & Sфrensen, T. I. (1994). Parental neglect during childhood and increased risk of obesity in young adulthood. Lancet, 34, 324-327. Livingstone, B. (2000). Epidemiology of childhood obesity in Europe. European Journal  of Pediatrics, 159 (Suppl.), 14-34. Lloyd, K., & Devine, P. (2006). Parenting practices in Northern Ireland: Evidence from the Northern Ireland Household Survey. Child Care in Practice, 12, 365-376. Maccoby, E. (2007). Historical overview of socialization research and theory. In J. E.  200 Grusec & P. D. Hastings (Eds.). Handbook of socialization: Theory and research (pp. 13- 41). New York: Guilford Press. Maccoby, E., & Martin, J. (1983). Socialization in the context of the family: Parent-child interaction. In P. H. Mussen (Series Ed.) & E. M. Hetherington (Vol. Ed.), Handbook of child psychology: Vol. 4. Socialization, personality, and social development (4th ed., pp. 1-101). New York: Wiley. Magarey, A., Daniels, L., & Boutlon, J. (2001). Prevalence of overweight and obesity in Australian children and adolescents: Reassessment of 1985 and 1955 data against new standard international definitions. Medical Journal of Australia, 174, 561-564. Mangelsdorf, S., Gunner, M., Kestenbaum, R., Lang, S., & Andreas, D. (1990). Infant proneness-to-distress temperament, maternal personality, and mother-infant attachment: Associations and goodness of fit. Child Development, 61, 820-831. Marston, A. R., London, P., & Cooper, L. M. (1976). A note on the eating behavior of children varying in weight. Journal of Child Psychology & Psychiatry, 17, 221-224. Martin, G. C., Wertheim, E. H., Prior, M., Smart, D., Sanson, A., & Oberklaid, F. (2000). A longitudinal study of the role of childhood temperament in the later development of eating concerns. International Journal of Eating Disorders, 27, 150-162. Martorell, R., Khan, L., Hughes, M., & Grummer-Strawn, L. (2000). Overweight and obesity in preschool children from developing countries. International Journal of Obesity, 24, 959-967. Maziade, M., Caronn, C., Cote, R., Merrett, C., Bernier, H., Laplante, B, et al. (1990). Psychiatric status of adolescents who had extreme temperaments at age 7. American Journal of Psychiatry, 147, 1531-1536.  201 Mehrabian, A., & Riccioni, M. (1986). Measures of eating-related characteristics for the general population: Relationships with temperament. Journal of Personality Assessment, 50, 610-629. Mei, Z., Scanlon, K., Grummer-Strawn, L., Freedman, D., Yip, R., & Trowbridge, F. (1998). Increasing prevalence of overweight among US low-income preschool children: The Centers for Disease Control and Prevention’s Pediatric Nutrition Surveillance 1983-1995 [Electronic version]. Pediatrics, 101, E12. Miles, J., & Shevlin, M. (2004). Applying regression and correlation: A guide forstudents and researcher. Thousand Oaks, CA: Sage. Morris, A., Silk, J., Steinberg, L., Sessa, F., Avenevoli, S., & Essex, M. (2002). Temperamental vulnerability and negative parenting as interacting predictors of child adjustment. Journal of Marriage and Family, 64, 461-471.  Mo-Suwan, L., Lebel, L., Puetpaiboon, A., & Junjana, C. (1999). School performance and weight status of children and young adolescents in a transitional society in Thailand. International Journal of Obesity Related Metabolic Disorder, 23, 272-277. Must, A. (1996). Morbidity and mortality associated with elevated weight in children and adolescents. American Journal of Clinical Nutrition, 63(Suppl.), 445-447. Must, A., Dallal, G. E., & Dietz, W.H. (1991). Reference data for obesity: 85th and 95th percentiles of body mass index (wt/ht²) and triceps skinfold thickness. American Journal of Clinical Nutrition, 53, 839-846. Must, A., & Strauss, R. S. (1999). Risks and consequences of childhood and adolescent obesity. International Journal of Obesity, 23(Suppl. 2), 2- 11.  Mustillo, S., Worthman, C., Erkanli, A., Keeler, G., Angold, A., & Costello, J. (2003).  202 Obesity and psychiatric disorder: Developmental trajectories. Pediatrics, 111, 851-859. Nagin, D. S. (1999). Analyzing developmental trajectories: A semi-parametric, group- based approach. Psychological Methods, 4, 139-157. Nagin, D. S. (2002). Overview of a semi-parametric, group-based approach for analyzing trajectories of development. Proceedings of Statistics Canada Symposium: Modeling survey data for social and economic research. Retrieved  December, 13, 2004, from www.statcan.ca/english/freepub/ 11-522-XIE/2002001/session10/nagin.pdf. Nagin, D. S. (2005). Group-based modeling of development. Cambridge, MA: Harvard University Press. Nagin, D. S., Pagani, R., Tremblay, R. E., & Vitaro, F. (2003). Life course turning points: a case study of effect of school failure on interpersonal violence. Development and Psychopathology, 15, 343-361. Nagin, D. S., & Tremblay, R. E. (2001). Analyzing developmental trajectories of distinct but related behaviors: A group-based method. Psychological Methods, 6, 18-34. Newman, W. P., Freedman, D. S., Voors, A. W., Gard, P. D., Srinivasan, S. R., Cresanta, J. L., et al. (1986). Relation of serum lipoprotein levels and systolic blood pressure to early atherosclerosis: The Bogalusa Heart Study. New England Journal of Medicine, 314, 138- 144. Nicklas, T. A., Baranowski, T., Baranowski, J. C., Cullen, K., Rittenberry, L., & Olvera, N. (2001). Family and child-care provider influences on preschool children’s fruit, juice, and vegetable consumption. Nutrition Reviews, 59, 224-235.  203 Niedhammer, I., Bugel, I., Bonenfant, S., Goldberg, M., & Lederc, A. (2000). Validity of self-reported weight and height in the French GAZEL cohort. International Journal of Obesity, 24, 1111-1118. O’Dea, J., & Caputi, P. (2001). Association between socioeconomic status, weight, age, and gender, and the body image and weight control practices of 6- to 19-year-old children and adolescents. Health Education Research, 16, 521-532. Ogden, C .L. (2004). Defining overweight in children using growth charts. Maryland Medicine, 5(3), 19-21. Ogden, C. L., Kuczmarski, R. J., Flegal, K. M., et al. (2001). Centers for Disease Control 2000 growth charts for the United States: Improvements to the 1977 National Center for Health Statistics version. Pediatrics, 109, 45-60. Ogden, C. L., Troiano, R., Briefel, R., Kuczmarski, R., Flegal, K., & Johnson, C. (1997). Prevalence of overweight among preschool children  in the United States, 1971 through 1994 [Electronic version]. Pediatrics, 99, E1. O’Loughlin, J., Paradis, G., Meshefedjian, G., & Gray-Donald, K. (2000). A five year trend of increasing obesity among elementary schoolchildren in multiethnic, low-income, inner-city neighborhoods in Montreal, Canada. International Journal of Obesity, 24, 1176-1182. Olson, S. L., Bates, J. E., & Bayles, K. (1982). Maternal perceptions of infant and toddler behavior: A longitudinal, construct validation study. Infant Behavior and Development, 5, 397-410.  204 Olvera-Ezzell, N., Power, T., & Cousins, J. (1990). Maternal socialization of children’s eating habits: Strategies used by obese Mexican-American children. Child Development, 61, 395-400. Parsons T. J., Power, C., Logan, S., & Summerbell, C. D. (1999). Childhood predictors of adult obesity: A systematic review. International Journal of Obesity, 23(Suppl. 8), 1-107. Pettit, G. S., & Bates, J. E. (1989). Family interaction patterns and children’s behavior problems from infancy to age 4 years. Developmental Psychology, 25, 413-420. Pietrobelli, A., Faith, M. S., Allison, D. B., Gallagher, D., Chiumello, G., & Heymsfeld, S. B. (1998). Body mass index as a measure of adiposity among children and adolescents: A validity study. Journal of Pediatrics, 132, 204-214. Power, C., Lake, J. K., & Cole, T. J. (1997). Body mass index and height from childhood to adulthood in the 1958 British born cohort. American Journal of Clinical Nutrition, 66, 1094-1101. Räikkönen, K., & Keltikangas-Järvinen, L. (1991). Hostility and its association with behavioral and somatic coronary risk indicators in Finnish adolescents and young adults. Social Science and Medicine, 10, 1171-1178. Räikkönen, K., Matthews, K., & Salomon, K. (2003). Hostility predicts metabolic syndrome risk factors in children and adolescents. Health Psychology, 22, 279-286. Ravaja, M. A., & Keltikangas-Järvinen, L. (1995). Temperament and metabolic syndrome precursors in children: A three-year follow up. Preventive Medicine, 24, 518- 537. Ravaja, M. A., Keltikangas-Järvinen, L., & Keskivaara, P. (1996). Type A factors as predictors of changes in the metabolic syndrome precursors in adolescents and young  205 adults: A 3-year follow-up study. Health Psychology, 15, 18-29. Ravelli, G. P., Slein, Z. A., & Susser, M. W. (1976). Obesity in young men after famine exposure in utero and early infancy. New England Journal of Medicine, 295, 349-353. Ravens-Sieberer, U., Redegeld, M., & Bullinger, M. (2001). Quality of life after in- patient rehabilitation in children with obesity. International Journal of Obesity, 25(Suppl.), 63-65. Reilly, J., Methven, E., McDowell, Z., Hacking, B., Alexander, D., Stewart, C., et al. (2003). Health consequences of obesity. Archives of Disease in Childhood, 88, 748-752. Rhee, K., Lumeng, J., Appugliese, D.,  Kaciroti, N., & Bradley, R. (2006). Parenting styles and overweight status in first grade. Pediatrics, 117, 2047-2054. Richardson, S. A., Goodman, N., Hastorf, A. H., & Dornbusch, S. M. (1961). Cultural  uniformity in reaction to physical disabilities. American Society Review, 1, 241-247. Richman, N., Stevenson, J., & Graham, P. J. (1982). Preschool to school: A behavioral study. London: Academic Press. Roberts, R. (1995). Can self-reported data accurately describe the prevalence of overweight? Public Health, 109, 275-284. Robinson, T. N. (1999). Reducing children’s television viewing to prevent obesity: A  randomized controlled trial. JAMA, 282, 1561-1567. Robinson, T. N., Kiernan, M., Matheson, D. M., & Haydel, K. F. (2001). Is parental control over children’s eating associated with childhood obesity? Results from a population- based sample of third graders. Obesity Research, 9, 306-312.  206 Rolland-Cachera, M. F., Deheeger, M., Bellisle, F., Sempe, M., Guilloud-Bataille, M., & Patois, E. (1984). Adiposity rebound in children: A simple indicator for predicting obesity. American Journal of Clinical Nutrition, 39, 129-135. Rosenbaum, M., Leibel, R. L., & Hirsch, J. (1998). Obesity. New England Journal of  Medicine, 338, 64-65. Rothbart, M. K. (1986). Longitudinal observation of infant temperament. Developmental Psychology, 22, 356-365. Rothbart, M. K. (1987). A psychobiologic approach to the study of temperament. In G. Kohnstamm & J. Bates (Eds.), Temperament discussed (pp. 63-77). Lisse: Swets & Zeitlinger. Rothbart, M. K. (1989). Biological processes in temperament. In G. A. Kohnstamm, J. E.  Bates, & M. K. Rothbart (Eds.), Temperament in childhood (pp. 76-110). Chichester, UK: John Wiley. Rothbart, M. K., & Ahadi, S. A. (1994). Temperament and the development of personality. Journal of Abnormal Psychology, 103, 55-66. Rothbart, M. K., Ahadi, S. A., & Hershey, K. L. (1994). Temperament and social behavior in childhood. Merrill-Palmer Quarterly, 40(1), 21-39. Rothbart, M. K., & Bates, J. E. (1998). Temperament. In W. Damon (Series Ed.) & N. Eisenberg (Vol. Ed.), Handbook of childhood psychology: Vol. 3. Social, emotional and personality development (pp. 105-176). New York: John Wiley. Rothbart, M. K., & Derryberry, D. (1981). Development of individual differences in temperament. In M. E. Lamb & A. L. Brown (Eds.), Advances in developmental Psychology (Vol. 1, pp. 37-86). Hillsdale, NJ: Erlbaum.  207 Rubin, K., Hastings, P., Chen, X., Stewart, S., & McNicholl, K. (1998). Interpersonal and maternal correlates of aggression, conflict, and externalizing problems in toddlers. Child Development, 68, 1614-1629. Russell, J. D., Kopec-Schrader, E., Rey, J. M., & Beumont, P. J. (1992). The parental bonding instrument in adolescent patients with anorexia nervosa. Acta Psychiatrica Scandinavica, 86, 236-239. Saelens, B. E., Ernst, M. M., & Epstein, L. H. (2000). Maternal child feeding practices and obesity: A discordant sibling analysis. International Journal of Eating Disorders, 27, 459-463. Sallis, J. F., Broyles, S. L., Frank-Spoh Ogden, G., Berry, C., Davis, T., & Nader, P. (1995). Children’s home environment in relation to mother’s adiposity. International Journal of Obesity and Related Metabolic Disorders, 19, 190-197. Sanci, L., Glover, S., & Coffey, C. (2003). Adolescent health education programmes: Theoretical principles in design and delivery. Annals of Academy of Medicine Singapore, 32(1), 78-85. Sanson, A., Hemphill, S., & Smart, D. (2002). Temperament and social development. In P. K. Smith & C. H. Hart (Eds.), Handbook of childhood social development (pp. 97- 116). London, UK: Blackwell. Sanson, A., Hemphill, S., & Smart, D. (2004). Connection between temperament and social development: A review. Social Development, 13, 142-170. Sanson, A., & Wise, S. (2001). Children and parenting: The past hundred years. Family Matter, 6, 36-45. Sanson, A. V., & Rothbart, M. K. (1995). Child temperament and parenting. In M.  208 Bornstein (Ed.), Handbook of parenting: Vol. 4. Applied and practical parenting (pp. 299-321). New York: Erlbaum. Scarr, S., & McCarthy, K. (1983). How people make their own environments: A theory of genotype-environment effects. Child Development, 54, 434-435. Schaffer, H. R. (1996). Social development: An introduction. Cambridge, MA: Blackwell. Schaffer, H. R. (1999). Understanding socialization: From unidirectional to bidirectional conceptions. In M. Bennet (Ed.), Developmental psychology (pp. 272-288). Philadelphia: Psychology Press. Schmitz, S., Saudino, K. J., Plomin, R., Fulker, D. W., & DeFries, J. C. (1996). Genetics and environmental influences on temperament in middle childhood: Analyses of teacher and tester ratings. Child Development, 67, 409-422. Schumann, L., Nichols, M. R., & Livingston, D. (2002). Preventing pediatric obesity: Assessment and management in the primary care setting. Journal of the American Academy of Nurse Practitioners, 14(2), 55-62. Schwarz, G. (1978). Estimating dimensions of a model. Annals of Statistics, 6, 461- 464. Schwimmer, J. B. (2003). Childhood obesity: The cases for coverage. Harvard Health Policy Review, 4(2), 62-72. Schwimmer, J., Burwinkle, T., & Varni, J. (2003). Health-related quality of life of severely obese children and adolescents. JAMA, 289, 1813-1819. Sekine, M., Yamagami,T., Hamanishi, S, & Kagamimori, S. (2002). Accuracy of the estimated prevalence of childhood obesity from height and weight values reported parents: Results of the Toyama Birth Cohort study. Journal of Epidemiology, 112, 9-13. Shephard, R. J. (2004). Role of the physician in childhood obesity. Clinical Journal of  209 Sport and Medicine, 14, 161-168. Sherman, J. B., Alexander, M. A., Dean, A. H., & Kim, M. (1995). Obesity in Mexican- American and Anglo children. Progressive Cardiovascular Nursing, 10, 27-34. Siervogal, R. M., Roche, A. F., Guo, S., Mukherjee, D., & Cameron, C. (1991). Patterns of change in weight²/stature from 2 to 18 years: findings from long-term serial data for children in Fels longitudinal growth study. International Journal of Obesity, 15, 479- 485. Sobal, J., & Stunkard, A. J. (1989). Socioeconomic status and obesity: A review of  the literature. Psychological Bulletin, 105, 260-275. Someya, T., Uehara, T., Kadowaki, M., Tangs, S., & Takahashi, S. (2000). Effects of gender difference and birth order on perceived parenting styles, measured by the EMBU scale, in Japanese two-sibling subjects. Psychiatry and Clinical Neurosciences, 54, 77- 81. Sothern, M. S., & Gordon, S. T. (2003). Prevention of obesity in young children: A critical challenge for medical professionals. Clinical Pediatrics, 42, 101-111. Spruijt-Metz, D., Lindquist, C., Birch, L., Fisher, J., & Goran, M. (2002). Relations between mothers’ child-feeding practices and children’s adiposity. American Journal of Clinical Nutrition, 75, 581-586. Statistics Canada. (2002, October 18). National Longitudinal Survey of Children and Youth: Childhood obesity: 1994-1999. The Daily, 6-7. Statistics Canada. The research data centre program. Retrieved November 13, 2007, from http://www.statcan.ca/english/rdc/index.htm   210 Steinberg, L. (2000). The family at adolescence: Transition and transformation. Journal of Adolescent Health, 27, 170-178. Stettler, N., Zemel, B. S., Kumanyika, S., & Stallings, V. A. (2002). Infant weight gain and childhood overweight status in a multi-center cohort study. Pediatrics, 109, 194-199. Stoolmiller, M. (1995). Using latent growth-curve models to study developmental processes. In J. M. Gottman (Ed.), The analysis of change (pp. 103-138). Mahwah, NJ: Erlbaum. Story, M. T., Neumark-Stzainer, D. R., Sherwood, N. E., Holt, K., Sofka, D., Trowbridge, F. L., et al. (2002). Management of child and adolescent obesity: Attitudes, barriers, skills, and training needs among health care professionals. Pediatrics, 110, 210-214. Strauss, R. S., & Knight, J. (1999). Influence of the home environment on the development of obesity in children. Pediatrics, 103, 85-93. Strayhorn, J. M., & Weidman, C. S. (1988). A parent practices scale and its relation to parent and child mental health. Journal of the American Academy of Child & Adolescent Psychiatry, 27, 613-618. Strelau, J. (1985). Introduction. In Temperamental bases of behavior: Warsaw studies on individual differences (pp. 1- 6). Lisse, Netherlands: Swets & Zeitlinger. Stunkard, A. J., & Sorensen, T. I. (1993). Obesity and socioeconomic status: A complex relation. New England Journal of Medicine, 329, 1036-1037. Summerbell, C. D., Ashton, V., Campbell, K. J., Edmunds, L., Kelly, S., & Waters, E. (2004). Interventions for treating obesity in children. The Cochrane Library, 3, CD001871. Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics. Boston: Pearson/  211 Allyn & Bacon. Taras, H. L., Sallis, J. F., Nader P. R., & Nelson, J. (1990). Children’s television-viewing habits and the family environment. American Journal of Diseases of Children, 144, 353- 359. Teachman, B. A., & Brownell, K. D. (2001). Implicit anti-fat bias among health professionals: Is anyone immune? International Journal of Obesity, 25, 1525-1531. Thomas, A., & Chess, S. (1977). Temperament and development. Brunner-Mazel: New York. Thomas, A., & Chess, S. (1982). Temperament and development. Brunner-Mazel: New York. Thomas, A., Chess, S., & Birch, H. G. (1968). Temperament and behavior disorders in children. New York: New York University Press. Thorpe, L., List, D., Marx, T., May, L., Helgeron, S., & Frieden, T. (2004). Childhood obesity in New York City elementary school students. American Journal of Public Health, 94(9), 1496-1500. Tinsley, B. J. (1992). Multiple influences on the acquisition and socialization of children’s health attitudes and behavior: An integrated review. Child Development, 63, 1043-1069. Tinsley, B. J. (2003). How children learn to be healthy. Cambridge, UK: Cambridge University Press. Tremblay, M. S., Katzmarzky, P. T., & Willms, J. D. (2002). Temporal trends in overweight and obesity in Canada, 1981-1996. International Journal of Obesity, 26, 538- 543. Tremblay, M. S., & Willms, J. D. (2000). Secular trends in the body mass index of Canadian children. Canadian Medical Association Journal, 163, 1429-1433. Tremblay, M. S., & Willms, J. D. (2003). Is the Canadian childhood obesity epidemic  212 related to physical inactivity? International Journal of Obesity, 27, 1100-1105. Troiano, R. P., & Flegal, K. M. (1998). Overweight children and adolescents: Description, epidemiology, and demographics. Pediatrics, 101, 497-504. Troiano, R. P., & Flegal, K. M. (1999). Overweight prevalence among youth in the US: Why so many different numbers? International Journal of Obesity and Metabolic Disorders, 23(Suppl. 3), 22-27. Troiano, R. P., Flegal, K. M., Kuczmarsky, J., Campbell, S. M., & Johnson, C. L. (1995). Overweight prevalence and trends for children and adolescents: The National Health and Nutrition Examination Surveys, 1963-1991. Archives of Pediatrics and Adolescent Medicine, 149, 1085-1091. Troiano, R. P., Frongillo, E. A., Sobal, J., & Levitsky, D.A. (1996). The relationship between body weight and mortality: A quantitative analysis of combined information from existing studies. International Journal of Obesity and Related Metabolic Disorders, 20, 63-75. Trost, S. G., Pate, R. R., Dowda, M., Saunders, R., Ward, D. S., & Felton, G. (1996). Gender differences in physical activity and determinants of physical activity in rural fifth grade children. Journal of School Health, 66, 145-151. Viner, R. M., Segal, T. Y., Lichtarowicz-Krynska, E., & Hindmarsh, P. (2005). Prevalence of insulin resistance in obesity. Archives of Disease in Childhood, 90, 10-14. Vitaro, F., Barker, E., Boivin, M., Brendgen, M., & Tremblay, R. (2006). Do early temperament and harsh parenting differentially predict reactive and proactive aggression? Journal of Abnormal Child Psychology, 34, 685-695.  213 Wachs, T., & Kohnstamm, G. (2000). The bidirectional nature of temperament-context links. In T. Wachs & G. Kohnstamm (Eds.), Temperament in context (pp. 201-222). New Jersey: Lawrence Erlbaum. Wang, Y. (2001). Cross-national comparison of childhood obesity: The epidemic and the relationship between obesity and socioeconomic status. International Journal of Epidemiology, 30, 1129-1136. Wang, Y., Monteiro, C., & Popkin, P. M. (2002). Trends of obesity and underweight in older children and adolescents in the US, Brazil, China and Russia. American Journal of Clinical Nutrition, 75, 971-977. Wardle, J., Sanderson, S., Guthrie, C. A., Rapoport, L., & Plomin, R. (2002). Parental feeding style and the inter-generational transmission of obesity risk. Obesity Research, 19, 453-462. Warren, J. M., Henry, C. J., Lightowler, S. M., & Perwaiz, S. (2003). Evaluation of a pilot school programme aimed at the prevention of obesity in children. Health Promotion International, 18, 287-296. Warren, S., & Simmens, S. (2000). Predicting toddler anxiety/depressive symptoms: Effects of caregiver sensitivity on temperamentally vulnerable children. Infant Mental Health Journal, 26, 40-55. Waxman, M., & Stunkard, A. J. (1980). Caloric intake and expenditure of obese boys. Journal of Pediatrics, 96, 187-193. Weicha, J., & Casey, V. (1994). High prevalence of overweight and short stature among Head Start children in Massachusetts. Public Health Reports, 109, 767-773.  214 Wells, J. C., Cole, T. J., & Davies, P. S. (1996). Total energy expenditure and body composition in early infancy. Archives of Disease in Childhood, 75, 423-426. Wells, J. C., & Davies, P. S. (1995). Diet and behavioral activity in 12-week-old infants. Annals of Human Biology, 22, 207-215. Wells, J. C., Stanley, M., Laidlaw, A. S., Day, J. M., Stafford, M., & Davies, P. S. (1997). Investigation of the relationship between infant temperament and later body composition. International Journal of Obesity, 21, 400-406. Whitaker, R. C., & Dietz, W. H. (1998). Role of parental environment in the development of obesity. Journal of Pediatrics, 132, 768-779. Whitaker, R. C., Pepe, M. S., Wright, J. A., Seidel, K. D., & Dietz, W.H. (1998). Early adiposity rebound and the risk of adult obesity [Electronic version]. Pediatrics, 10(3), E5. Whitaker, R. C., Wright, J. A., Pepe, M. S., Seidel, K. D., & Dietz, W. H. (1997). Predicting obesity in young adulthood from childhood and parental obesity. New England Journal of Medicine, 337, 869-873. Wilkins, S. C., Kendrick, O. W., Stitt, K. R., Stinett, N., & Hammarlund, V.A. (1998). Family functioning is related to overweight in children. Journal of the American Dietetic Association, 5, 572-574. Willms, J. D. (2002). A study of vulnerable children. In J. D. Willms (Ed.), Vulnerable children: Findings from Canada’s National Longitudinal Survey of Children & Youth (pp. 3-22). Edmonton, AB: University of Alberta Press. Willms, J. D., Tremblay, M. S., & Katzmarzky, P. T. (2003). Geographic and demographic variation in the prevalence of overweight Canadian children. Obesity Research, 11, 668-673.  215 Wills, T. A., Cleary, S. D., Shiner, O., & Fegan, M. (2002). Temperament dimensions and health behavior: A developmental model. In L. L. Hayman, M. M. Mahon, & J. R. Turner (Eds.), Health and behavior in childhood and adolescence (pp. 3-36). New York: Springer. Wolkind, S., & de Salis, W. (1982). Infant temperament, maternal mental state and child behavioral problems. In R. Porter & G. Collins (Eds.), Temperamental differences in infants and young children. London: Plenum. World Health Organization. (2000). Obesity: Prevention and managing the global epidemic. Geneva, Switzerland: Author. WHO Multicenter Growth Reference Study Group. (2006). WHO child growth standards: length/height-for-age, weight-for-age, weight-for-length, weight-for-height, and body mass index-for-age methods and development. Department of Nutrition for Health and Development. Geneva, Switzerland: Author. Zhou, Q., Eisenberg, N., Losoya, S., Fabes, R., Reiser, M., Guthrie, I., et al. (2002). The relations of parental warmth and positive expressiveness to children’s empathy-related responding and social functioning: a longitudinal study. Child Development, 73, 893-915. Zigler, E. (1995). Forward. In M. H. Bornstein (Ed.), Handbook of parenting: Vol. 2. Biology and ecology of parenting (pp. xi-xii). New York: Lawrence Erlbaum.  216 Appendix A  The Research Data Centres Program (from: www.statcan.ca/english/rdc/index.htm) “The Research Data Centres (RDC) program is part of an initiative by Statistics Canada, the Social Sciences and Humanities Research Council (SSHRC) and university consortia to help strengthen Canada's social research capacity and to support the policy research community.” “RDCs provide researchers with access, in a secure university setting, to microdata from population and household surveys. The centres are staffed by Statistics Canada employees. They are operated under the provisions of the Statistics Act in accordance with all the confidentiality rules and are accessible only to researchers with approved projects who have been sworn in under the Statistics Act as 'deemed employees.'” “In 1998, the Canadian Initiative on Social Statistics studied the challenges facing the research community in Canada. One of the recommendations of the national task force report on the Advancement of Research using Social Statistics was the creation of research facilities to give academic researchers improved access to Statistics Canada's microdata files. This access would allow researchers in the social sciences to build expertise in quantitative methodology and analysis.” “To access the microdata housed in the Research Data Centres (RDCs), researchers submit a project proposal to an adjudicating committee operating under the auspices of the Social Sciences and Humanities Research Council (SSHRC) and Statistics Canada.”  217 “The university-based centres are, essentially, extensions of Statistics Canada offices, with a full-time Statistics Canada employee at each site to screen the results that will be released and ensure compliance with confidentiality policies and procedures. The centres operate under the same security provisions as any other Statistics Canada offices, including the use of physical access controls and stand-alone computers with no links outside Statistics Canada.”  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0066157/manifest

Comment

Related Items