Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Measurement invariance of body image across the adult life span : can we compare across age and gender… Rusticus, Shayna Ann 2005

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

Item Metadata

Download

Media
831-ubc_2005-0622.pdf [ 5.73MB ]
Metadata
JSON: 831-1.0054394.json
JSON-LD: 831-1.0054394-ld.json
RDF/XML (Pretty): 831-1.0054394-rdf.xml
RDF/JSON: 831-1.0054394-rdf.json
Turtle: 831-1.0054394-turtle.txt
N-Triples: 831-1.0054394-rdf-ntriples.txt
Original Record: 831-1.0054394-source.json
Full Text
831-1.0054394-fulltext.txt
Citation
831-1.0054394.ris

Full Text

MEASUREMENT INVARIANCE OF BODY IMAGE ACROSS THE ADULT LIFE SPAN: CAN WE COMPARE ACROSS AGE AND GENDER WITH BODY IMAGE MEASURES? by SHAYNA ANN RUSTICUS B. A . , Kwantlen University College, 2003 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF M A S T E R OF ARTS In THE FACULTY OF GRADUATE STUDIES Measurement, Evaluation, and Research Methodology THE UNIVERSITY OF BRITISH COLUMBIA September, 2005 © Shayna Ann Rusticus, 2005 Body Image i i Abstract The issue of body image has been widely discussed in the literature as it pertains to adolescents and young adults; however, body image issues among older individuals, and especially among older men, have been largely ignored. Many of the instruments used to measure the theoretical construct of body image have largely been developed with younger populations. However, before these instruments can be applied to older populations, they must exhibit adequate cross-group equivalence. The hypotheses that the Multidimensional Body-Self Relations Questionnaire (MBSRQ), the Appearance Schemas Inventory-Revised (ASI-R), and the Body Image Quality of Life Inventory (BIQLI) can be used to make cross-age and gender comparisons was examined in a sample of 422 men (185 young, 131 middle-aged, 106 older) and 840 women (364 young, 267 middle-aged, 209 older). The results of the measurement invariance tests (i.e., configural, metric, and scalar) for the subscales of the MBSRQ clearly illustrate that the multidimensional nature of body image is perceived quite differently across the age and gender groups and thus the applicability of these subscales depends on both the characteristics of the sample and the goals of the study. The results for the ASI-R revealed that all groups, except for the older women, met requirements for all three levels of invariance tested and that comparisons may be conducted across age group for men and across gender for young and middle-aged adults. Results for the BIQLI indicated that comparisons may be conducted across all age and gender groups. The inferences that may be made at each level of configural, metric, and scalar invariance are highlighted. For those scales exhibiting scalar invariance, appropriate comparisons are conducted and discussed. Body Image i i i T A B L E OF C O N T E N T S Abstract 1 1 Table o f Contents i« List of Tables v List of Figures • • v u Acknowledgments viii CHAPTER I Introduction , 1 CHAPTER II Review of the Literature 5 General Body Image Concerns in Men and Women 5 Gender Differences in Body Image Between Young Men and Women... 9 Body Image Issues in Older Women 12 Gender Similarities and Differences in Body Image in Middle-Aged And Older Adults , 17 Measurement Invariance in Cross-Group Research 24 Present Study 29 CHAPTER III Method 33 Participants 33 Measures 35 Procedure 37 Model Evaluation 39 CHAPTER IV Results 46 Missing and Ambiguous Data 46 Body Image iv Invariance of the Multidimensional Body-Self Relations Questionnaire 52 Gender and Age Differences in Mean Scores on the MBSRQ Subscales 82 Invariance of the Appearance Schemas Inventory-Revised 85 Gender and Age Differences in Mean Scores on the ASI-R Subscales... 91 Invariance of the Body Image Quality of Life Inventory 92 Gender and Age Differences in Mean Scores on the BIQLI 98 CHAPTER V Discussion 101 Multidimensional Body-Self Relations Questionnaire 103 Appearance Schemas Inventory-Revised 108 Body Image Quality of Life Inventory 110 Conclusion I l l Strengths of the Study 113 Limitations 114 Future Directions 115 References 117 Body Image v List of Tables Number Title Page Table 1 Demographic Characteristics of Participants by Gender and Age Group (N= 1262) 34 Table 2 Absolute Skewness, Kurtosis, and Multivariate Kurtosis Values for the MBSRQ, ASI-R, and BIQLI 50 Table 3 Goodness-of-Fit Indices for the M B S R Q - Appearance Evaluation 53 Table 4 Tests of Measurement Invariance for the MBSRQ - Appearance Evaluation.. 54 Table 5 Goodness-of-Fit Indices for the M B S R Q - Appearance Orientation 57 Table 6 Tests of Measurement Invariance for the MBSRQ - Appearance Orientation.. 58 Table 7 Goodness-of-Fit Indices for the M B S R Q - Fitness Orientation 60 Table 8 Tests of Measurement Invariance for the MBSRQ - Fitness Orientation 61 Table 9 Goodness-of-Fit Indices for the M B S R Q - Health Evaluation .. . .: 63 Table 10 Tests of Measurement Invariance for the MBSRQ -Health Evaluation 64 Table 11 Goodness-of-Fit Indices for the M B S R Q - Health Orientation 68 Table 12 Tests of Measurement Invariance for the MBSRQ -Health Orientation 69 Table 13 Goodness-of-Fit Indices for the M B S R Q - Illness Orientation 72 Table 14 Tests of Measurement Invariance for the MBSRQ -Illness Orientation 73 Table 15 Goodness-of-Fit Indices for the MBSRQ - Body Areas Satisfaction 75 Table 16 Tests of Measurement Invariance for the MBSRQ - Body Areas Satisfaction. 76 Table 17 Goodness-of-Fit Indices for the MBSRQ - Overweight Preoccupation 79 Table 18 Tests of Measurement Invariance for the MBSRQ - Overweight Preoccupation 80 Table 19 Means (Standard Deviations) for the Subscales of the MBSRQ 82 Body Image vi Table 20 Goodness-of-Fit Indices for the ASI-R Two-Factor Model 86 Table 21 Tests of Measurement Invariance for the ASI-R Two-Factor Model 87 Table 22 Goodness-of-Fit Indices for the ASI-R One-Factor Model 90 Table 23 Means (Standard Deviations) for the Subscales of the ASI-R 91 Table 24 Goodness-of-Fit Indices for the BIQLI - Full Scale 93 Table 25 Tests of Measurement Invariance for the BIQLI 94 Table 26 Goodness-of-Fit Indices for the BIQLI - Reduced Scale 96 Table 27 Tests of Measurement Invariance for the Reduced BIQLI 97 Table 28 Means (Standard Deviations) for the Full BIQLI 99 Table 29 Means (Standard Deviations) for the Reduced BIQLI 100 Table 30 Levels of Invariance Attained for the Subscales of the MBSRQ 104 Body Image vii List of Figures Figure Title Page number Figure 1 Procedure for Assessing Measurement Invariance 43 Body Image viii Acknowledgments I would like to express my gratitude to my supervisor, Dr. Anita Hubley, whose expertise, guidance, and encouragement, added considerably to my graduate experience. I appreciate her vast knowledge and skill in many areas (e.g., body image, aging, ethics, interaction with participants and research assistants), and her assistance in writing reports (i.e., grant proposals, scholarship applications and this thesis). I would also like to acknowlege my committee members, Dr. Bruno Zumbo, Dr. Laurie Ford, and Dr. Kimberly Schonert-Reichl for the assistance they provided at all levels of the research project and their supportive and encouraging comments. Many thanks also go out to the volunteer research assistants, Anne Muscat, Sarah Chan, Beth Chan, and Kristine Hagen, who assisted me in the recruitment of the large number of participants I needed for this study. It was because of their help and dedication to the project that I was able to stay motivated and focused over the several months of recruitment. I would also like to thank my friends in the Adult Development and Psychometrics lab for their assistance in recruiting for this thesis, and for our exchanges of knowledge, skills, and venting of frustration during my graduate program, which helped enrich the experience. I would also like to acknowledge and thank many of my family member, friends, fellow students, and professors, who assisted in this thesis by helping to get the word out about the study. I would also like to thank my family for the support they provided me in this thesis and through my entire life, and in particular, I must acknowledge my husband and best friend, Marty, without whose love, encouragement, and support, I would not have finished this thesis. Finally, I would like to thank Dr. Thomas Cash for the use of his measures and his permission to exceed the number of copies normally allowed. In conclusion, I recognize that this research would not have been possible without the Body Image ix financial assistance of SSHRC, the University of British Columbia (Graduate Student Entrance Scholarship, UGF), and the Department of Educational and Counselling Psychology, and Special Education (Graduate Student Research Scholarship), and express my gratitude to those agencies. Body Image 1 Chapter One Introduction Body image has been defined as "the picture we have in our minds of the size, shape, and form of our bodies; and (refers) to our feelings concerning these characteristics and our constituent body parts" (Slade, 1988, p. 20; parenthetical statement added). The issue of body image has been widely discussed in the literature as it pertains to adolescents and young adults; however, body image issues among older individuals, and especially among older men, have been largely ignored. With over 47% of Canada's population being 40 years of age or older (Statistics Canada, 2004), there is a great need for more body image research with older populations. As our bodies develop and age, we cannot assume that older women and men will evaluate and perceive their bodies in the same ways as younger adults do, or that they perceive their own bodies the same way over the course of their life (Hurd, 2000). There are multitudes of physical signs that accompany the process of aging, such as the appearance of wrinkles and age spots, the thinning and graying of hair, the development of arthritis, or the need for devices such as hearing aids, eyeglasses, canes, or walkers (Chrisler & Ghiz, 1993), that most likely will have an impact upon body image (Kreuger, 1989). Although some body image measures have been used to assess similarities and differences in body image among men and women of different ages, an issue that has not been addressed in this literature is whether or not the measures used are invariant in all age and gender groups. A l l too often, measures are used with the assumption that they are measuring the same concept(s) across groups. However, when a measure is used to make comparisons between two or more groups, evidence is needed to show that the measure is functioning the same way in each group. In the body image literature, it is common for measures of body image to be validated for use with college samples and then used to make comparisons to other Body Image 2 age groups. There are two problems with this. First, validity evidence should be provided for each age group to demonstrate the validity of the measure for each group. Second, even i f the measure has been found to be valid in all age groups, it does not guarantee that the measure is invariant across groups for comparison purposes. Measurement invariance refers to whether "under different conditions of observing and studying phenomena, measurement operations yield measures of the same attribute" (Horn & McArdle, 1992, p. 117). Only through the demonstration of measurement invariance can a measure be deemed to be measuring the same attribute across groups. If there is no evidence of the presence or absence of measurement invariance, or i f invariance is not obtained, any differences found between groups cannot be interpreted unambiguously. For example, age differences on a scale measuring fitness importance might be due to true differences between age groups on the underlying construct, such that younger individuals rate their fitness levels as being more important to them than older individuals, or these differences might be due to systematic biases in the way people of different ages respond to certain items, such that an item asking about physical strength may be more salient than the other items to younger individuals whereas this is not the case for older individuals. Alternatively, findings of no differences between groups are also open to corresponding alternative interpretations. In short, as stated by Horn (1991), "Without evidence of measurement invariance, the conclusions of a study must be weak" (p. 119). Evidence of measurement invariance is accumulated on an incremental basis. The weakest form of invariance, configural invariance, assesses whether the configuration of the salient and nonsalient factor loadings is equivalent across groups. Configural invariance is the minimum condition for factorial invariance (Horn & McArdle, 1992). When configural invariance is met, it indicates that the same attribute is being measured in each of the groups. Body Image 3 The next level of invariance is metric invariance. Metric invariance assesses whether there is equality of the unstandardized factor pattern weights (i.e., factor loadings) across groups. This level of invariance provides a strong basis for inference that individuals from each of the groups interpret and respond to the measure in a similar way (Horn & McArdle, 1992). Finally, scalar invariance assesses whether there is consistency between group differences in latent means and group differences in observed means by examining the equality of the item intercepts. Evidence of scalar invariance is necessary in order to make mean comparisons across groups (Meredith, 1993; Steenkamp & Baumgartner, 1998). When scalar invariance is met, cross-group differences in the means of the observed items reflect differences in the means of the underlying construct(s). The body image measures chosen for the present study include the Multidimensional Body-Self Relations Questionnaire (MBSRQ), the Appearance Schemas Inventory-Revised (ASI-R), and the Body Image Quality of Life Inventory (BIQLI). For these measures to be used to assess the concept of body image across age or gender groups, it is necessary to demonstrate that these measures are invariant across the groups. Therefore, the purpose of this study is to examine the configural, metric, and scalar invariance of each of these measures across three age groups of men and women. Evidence of configural invariance will establish whether these measures can used to assess the concept of interest in each of the groups. Evidence of metric invariance will establish whether these measures can be used to examine structural relationships between the concept of interest and other concepts across the groups (Steenkamp & Baumgartner, 1998). And finally, evidence of scalar invariance will establish whether these measures are equivalent across the groups and can be used to make gender and age group comparisons. Several confirmatory factor analyses, performed at the multiple group level, will provide the basis for evaluating the cross-group applicability of these three scales. Body Image 4 First, it will be ascertained whether the same number of factors are identified by the same observed variables in each of the different samples (i.e., young men, young women, middle-aged men, middle-aged women, older men, older women). This represents a necessary preliminary step for testing generalizability across samples and establishes the equivalency of factor patterns, or configural invariance. This model is then used as a baseline model for examining further tests of invariance. Second, metric invariance or the invariance of factor loadings across samples will be investigated. Examination of factor loading invariance shows whether the measures indicate the same factors to the same extent across samples. Finally, scalar invariance will be tested by examining the fit of the model, where the vector of item intercepts is made invariant across samples. Evidence of scalar invariance would indicate that there is consistency between the latent means and the observed means, and that cross-group comparisons can be made validly. Body Image 5 Chapter Two Review of the Literature An important step in the extension of the body image literature to older populations and to men is to find measures that are measuring the construct of body image equally well for both men and women of different ages. Previous research that has used self-report body image measures appears not to have considered that there might be important differences in how men and women and older individuals conceptualize and perceive the construct of body image. Furthermore, the use of body image measures to examine differences among these groups without evidence that the measure is functioning equally well across the groups casts doubt on the validity of the interpretations and conclusions drawn. No research to date appears to have examined the measurement equivalence of any body image measure to determine i f these measures can be used to assess body image in men and women of different ages and can be used to make comparisons across groups. The present study will seek to fill this gap. First, I will review the current body image literature describing: (a) general body image concerns in men and women; (b) gender differences in body image between young adult men and women; (c) body image issues in older women; and, (d) gender similarities and differences in body image in older adults. Second, I will highlight the importance of examining measurement invariance in cross-group research and will summarize the literature on measurement invariance. Finally, I will outline the proposed study that will examine the equivalence of the MBSRQ, the ASI-R, and the BIQLI across gender and age. General Body Image Concerns in Men and Women Body image has been defined as "the picture we have in our minds of the size, shape, and form of our bodies; and (refers) to our feelings concerning these characteristics and our constituent body parts" (Slade, 1988, p. 20; parenthetical statement added). Over the last three Body Image 6 decades, a number of informal national surveys have been conducted in the U.S. questioning Americans about issues related to body image (Cash & Henry, 1995; Cash, Winstead, & Janda, 1986; Garner, 1997). Cash et al. (1986) followed up on a 1972 body image survey by Bercheid, Walster, and Bohrnstedt (1972, as cited in Cash et al., 1986) by investigating body image attitudes of nearly 30,000 American men and women aged 15 to 74, using the Body-Self Relations Questionnaire (BSRQ). Participants were placed into six age categories: younger than 20, 20 to 29, 30 to 39, 40 to 49, 50 to 59, and 60 to 74. Results showed that, overall, respondents in 1985 had higher levels of body dissatisfaction than respondents in 1972, and women were more dissatisfied than men in all areas except for face and height. The two groups of women in their teens and 20s were found to be the most dissatisfied with their appearance. Both women and men in these two age groups were also the most concerned with their appearance. This was followed by a steady decline of interest in appearance in those participants who were in their 30s, 40s, and 50s. Although the authors did not state i f it was significant, this pattern changed in the group of women in the 60 and above age category, who showed a slight rise in interest in their appearance. In another follow-up study conducted by Cash and Henry (1995), the authors examined body image attitudes among adult women in the United States using the Multidimensional Body-Self Relations Questionnaire (MBSRQ) and assessed any changes in body image from the national survey in done in 1985. Their sample consisted of 803 women categorized into five age groups: 18-24, 25-35, 35-44, 45-54, and 55-70. The findings indicated again that a large proportion of women of all ages reported a negative body image. Forty-five percent of the sample was specifically dissatisfied with their weight, lower torso, and mid torso. Only 16% of the women were satisfied with all physical aspects that were measured (i.e., face, height, hair, upper torso, muscle tone, weight, height, lower torso, and mid torso). The only Body Image 7 significant age effect found was that the 18 to 24 year olds had a more favorable body image relative to older groups. Compared to the 1985 results, there was an overall increase in negative global appearance evaluation. Garner (1997) conducted a fourth large-scale follow-up survey. This time, 4000 participants (3452 women and 548 men), ranging in age from 13 to 90, and grouped into decades, responded to the survey. Participants were asked to respond to numerous pages worth of questions, developed by the author, concerning how they saw, felt, and were influenced by their bodies. Consistent with the previous surveys, a large proportion of both women (56%) and men (43%) reported that they were dissatisfied with their overall appearance. In terms of specific body parts, women were most dissatisfied with their abdomen, weight, hips, and muscle tone and men were most dissatisfied with their abdomen, weight, muscle tone, and chest. It is interesting to note that, for both sexes, the stomach and weight were the top two major sources of disapproval. The results also indicated that women were more dissatisfied with their appearance than men at all ages. For women, levels of dissatisfaction remained fairly consistent across the age groups whereas for men there were peaks in dissatisfaction in their 30s and 50s. While a substantial percentage of both women (84%) and men (58%) reported dieting to lose weight, a prominent difference that surfaced between the genders was that a significant proportion of men also wanted to gain weight. This highlights an important distinction about the meaning of body dissatisfaction in men and women. For most women, body dissatisfaction results from a desire to lose weight whereas for most men body dissatisfaction is just as likely to result from men wanting to lose weight as from men wanting to gain weight. The cumulative results from the four body image surveys done in 1972, 1985, 1993, and 1997 provide evidence of increasingly negative body images among women and men of Body Image 8 all ages (Cash & Henry, 1995; Cash et al., 1986; Garner, 1997). In fact, there has been an increase in overall body dissatisfaction of 125% among women and nearly 200% among men between the 1972 and 1997 results (Garner, 1997). In a society in which the standards of beauty have become increasingly important in the lives of women and the ideal body has become increasingly thin and narrowly defined while women are, in fact, becoming larger, it should be of no surprise that women are developing increasingly more negative images of themselves (Wolf, 1991). In addition, with a recent rise in the valuation of the muscular male body and a surge in media pressure on men to be more concerned with their appearance and to conform to the male muscular ideal (Pope, Phillips, & Olivardia, 2000), it should also not be surprising that men too are developing poorer images of themselves. However, the finding that the older age groups did not report a poorer body image relative to the younger age groups is counter to the stereotype that suggests aging leads to a diminished body image. The authors explained these findings by suggesting that, as individuals' age, they adjust their standards appropriately (Cash et al., 1986). As a result, when women and men make social comparisons, they compare themselves to others their own age rather than to younger or "idealized" individuals, or to younger versions of themselves. In addition, it is hypothesized that the standards held by older generations may also be less harsh than those set by younger generations (Cash et al., 1986). Although these findings highlight some interesting trends among men and women and their perceptions of their body image, these findings must be interpreted with a certain degree of caution. The 1985, 1993, and 1997 studies each built on the conclusions drawn from the previous studies; however, these studies were conducted by different researchers, used different measures to assess body image, and used different samples. Thus, in addition to concerns about measurement invariance, one also must question whether these differences Body Image 9 come from using different measures and samples. That said, these findings should be interpreted as a general pattern of body image concerns. More rigorous research is needed to replicate these findings and trends. Another thing to consider, and also one of the major concerns of interpreting cross-sectional differences between age groups in most studies, is the inability to fully distinguish between age and cohort effects (Palmore, 1978). A l l too often, differences between age groups are assumed to be the result of the aging process and the fact that they may be partly or entirely the result of cohort differences is ignored. Age-related differences point to maturational changes in individuals whereas cohort-related differences refer to historical differences between groups of individuals (Marshall, 1983). Making the distinction between the two does not solve the problem as to what is at the root of any group differences, but, at the same time, failure to point out the competing explanation disguises the problem and is misleading. This point should be kept in mind while interpreting the results from the following studies that make claims about age-related differences. Gender Differences in Body Image Between Young Men and Women Concerns about physical appearance affect a substantial segment of the population and the most prominent body image concern is dissatisfaction with weight and body shape (Cash & Pruzinsky, 1990; Cash et al., 1986; Fallon & Rozin, 1985). As such, many of the studies investigating gender differences in body image have focused on self-ideal discrepancies regarding body size, shape, or weight (e.g., Fallon & Rozin, 1985; Jacobi & Cash, 1994; Lokken, Ferraro, Kirchner, & Bowling, 2003; Stanford & McCabe, 2002). Fallon and Rozin (1985) examined gender differences in perceptions of body shapes by asking 227 female and 248 male undergraduate students to complete a series of figure rating scales. For women, their current body shape was significantly bigger than both their ideal body Body Image 10 shape and the body shape they felt was most attractive to the opposite sex. For men, their current, ideal, and most attractive body shapes were nearly identical. However, further research has highlighted that self-ideal discrepancies for men are just as likely to result from wanting to be heavier as from wanting to be thinner (e.g., Jacobi & Cash, 1994). When one fails to take the sign of the discrepancy score into account, the mean ideal body shape preference for men tends to balance out somewhere near the current body shape, masking the levels of dissatisfaction men have with their bodies that are in opposite directions. Thus, researchers must take care to consider both the direction and magnitude of discrepancies. Lokken, Ferraro, Kirchner, and Bowling (2003) examined gender differences in body size dissatisfaction among individuals with low, medium, or high levels of body focus. Using a sample of 30 male and 30 female (10 each for low, medium, and high body focus) undergraduate students, the researchers had participants complete the Vocabulary subtest of the Wechsler Adult Intelligence Scale-Revised, the Beck Depression Inventory, and the Body Image Assessment Questionnaire. The results of a 2 (gender) x 3 (level of body focus) analysis of variance indicated a main effect for gender, with women reporting a significantly larger discrepancy between their real and ideal body shapes than men. However, like Fallon and Rozin (1985), the researchers failed to take into account the direction of the discrepancy for men, thus possibly obscuring differences in body dissatisfaction among men. There was no significant main effect for level of body focus and no significant interaction. Some caution should be used in interpreting the results of this study however as the small sample sizes resulted in relatively low power. Jacobi and Cash (1994) investigated discrepancies among multiple physical attributes -body size, weight, muscularity, height, hair length, hair color, eye color, and female breast size using a sample of 66 male and 69 female college students. For most of these attributes, Body Image 11 participants were asked to rate four percepts: actual self-percept, personal ideal, what the opposite sex would find ideal, and their ideal for the opposite sex. Results indicated that, for body size, 65% of men chose an ideal body size that was larger than their current size and 23% chose a smaller ideal. In addition, 91% of the men indicated that they would like to be more muscular, 62% wished to be taller, 55% wanted to weigh more than they currently do, and 42% wanted to weigh less. For women, 68% preferred an ideal body size that was smaller than their current size and only 10% selected a heavier ideal. The vast majority (93%) of women also wanted to weigh less than they currently do. Also, like the men, 78% of women wanted to be more muscular, and 46% would like to be taller (as opposed to only 3% who wanted to be shorter). Overall, the gender differences generally indicated that women desired to be lighter and thinner but more muscular, and men desired to be taller and more muscular. Stanford and McCabe (2002) examined discrepancies among males and females on various body parts. Sixty females and 50 males ranging in age from 18 to 22 completed Marsh's Self-Description Questionnaire, an Attractiveness/Effectiveness Questionnaire, and an Importance Scale. Participants also evaluated their body parts using a computerized body image program that enlarged or shrunk images of their actual body parts. Pertinent findings indicated that significant gender differences were evident for the ideal upper, middle, and lower body. Most men wanted an ideal upper body that was substantially larger whereas most women wanted an upper body that was moderately smaller. For the middle body, women wanted to be substantially smaller whereas there was a more even spread among the men. For the lower body, both men and women wanted to be smaller, but this was more so for the women. As can been seen from these studies, the general conclusions have been that both women and men are dissatisfied with their bodies. Most women want an ideal body shape that Body Image 12 is significantly smaller than their current body shape, and although most men also have a significant self-ideal discrepancy, they are as likely to want an ideal body shape that is larger as one that is smaller than their current body shape. Such body image concerns have been fairly well established with younger populations, with the vast majority of studies of this nature being conducted with college-aged students. Body image issues among older individuals, especially among older men, have been largely ignored. As our bodies develop and age, we cannot assume that older women and men will evaluate and perceive their bodies in the same ways as younger adults do, or that they perceive their own bodies the same way over the course of their life (Hurd, 2000). There are multitudes of physical signs that accompany the process of aging, such as the appearance of wrinkles and age spots, the thinning and graying of hair, the development of arthritis, or the need for devices such as hearing aids, eyeglasses, canes, or walkers (Chrisler & Ghiz, 1993), that most likely will have an impact upon body image (Kreuger, 1989). Body Image Issues in Older Women The majority of studies examining age-related perceptions of body image in older women have found that, although older women tend to see their bodies more negatively than younger women, they do not differ in levels of body part dissatisfaction compared to younger women (Deeks & McCabe, 2001; Hurd, 2000; Lewis & Cachelin, 2001; Tiggemann & Lynch, 2001; Webster & Tiggemann, 2003). Deeks and McCabe (2001) investigated differences . between pre-, peri-, and postmenopausal women's evaluations of appearance, health, fitness, preoccupation with weight, overall body satisfaction, and visual perception of body image using the MSBRQ and the Stunkard Body Shape Figure Scale. Their sample consisted of 304 Australian women ranging in age from 35 to 65. Although the authors did not find any effects related directly to menopause, they did find that the postmenopausal group, who, on average, Body Image 13 was older than the other two groups, was more negative about their appearance. There were no differences found between the groups on satisfaction with body parts. A l l three groups were equally more dissatisfied with lower and mid torso compared to upper torso, hair, face, and height. In looking at the results of the Body Shape Figures, it was found that ideal and attractive figures (i.e., what they thought the opposite sex would find most attractive) were smaller than their current shape. Although the figures chosen by both groups were relatively small, older women did choose progressively larger figures for ideal body shape. This suggests that women internalize societal expectations of smaller body shapes at any age, but there is some acceptance that older women will have a slightly larger shape (Deeks & McCabe, 2001). Lewis and Cachelin (2001) examined cohort differences in body image, drive for thinness, and eating attitudes among a middle-aged and elderly sample of women using the Body Mass Index, the Eating Disorder Inventory, Fear of Aging, and figure ratings. Their sample consisted of 250 community-dwelling women, 125 who were between the ages of 50 and 65, and 125 who were age 66 and above. Their results on the figure ratings were identical to those found by Deeks and McCabe (2001). For both groups, the women chose a current figure that was larger than both their ideal and attractive figure; however, the elderly group of women did choose a slightly larger, although still small, ideal figure compared to the middle-aged group. The authors also found a significantly higher discrepancy between current and ideal figures among the middle-aged group, resulting from the fact that the middle-aged women were more likely to choose a slightly smaller ideal figure. Also similar to the study by Deeks and McCabe (2001) was the finding that both groups had similar levels of body dissatisfaction. Tiggemann and Lynch (2001) set out to investigate the hypothesis that body dissatisfaction is stable across the life span through the assessment of several aspects of body Body Image 14 image in a broadly based sample of Australian women. Their sample consisted of 322 women, divided into six age categories: 20-29, 30-39, 40-49, 50-59, 60-69, and 70-85. The authors used a variety of measures, including Fallon and Rozin's Figure Rating Scale, the Body Esteem Scale, the Self-Objedification Scale, the Body Surveillance and Body Shame subscales of the Objectified Body Consciousness Scale, the short form of Appearance Anxiety Scale, the Drive for Thinness, Bulimia, and Body Dissatisfaction subscales of the Eating Disorder Inventory, the Revised Restraint Scale, and Bachman and O'Malley's (1977) revision of Rosenberg's Self-Esteem Scale, to fully examine the concept of body satisfaction and to investigate the processes underlying age differences by application of objectification theory. In brief, objectification theory argues that one consequence of being a woman in a culture that sexually objectifies the female body is that girls and women are gradually socialized to internalize an observer's perspective of their physical self and thus begin to treat themselves as an object to be looked at and evaluated. This "self-objectification" is characterized by habitual and constant monitoring of the body's outward appearance (Tiggemann & Lynch, 2001). Consistent with previous findings, Tiggemann and Lynch (2001) found that the women rated their ideal figure as smaller than their current figure, but both the current and ideal figures increased in size with age. There were no age-related differences found for body dissatisfaction or body esteem, indicating that these appear to remain stable across the life span. Of interest was the finding that self-objectification, habitual body monitoring, and appearance anxiety decreased with age, suggesting that, as women get older, the importance they place on the appearance of their body decreases (Tiggemann & Lynch, 2001). The authors do make note that one of the limitations of their study was the possibility that the obtained age differences might reflect historical changes rather than developmental changes that accompany the aging process. They state that, although the evidence does not seem to support this Body Image 15 possibility, it is possible that there were differences in the beauty stereotypes that women were subjected to at an earlier formative age that may have impacted the results of the study (Tiggemann & Lynch, 2001). Webster and Tiggemann (2003) explored the relationship between body dissatisfaction and the self across different age groups, while taking into account cognitive control (i.e., participants' perceived control over their bodies). Using a modified Body Cathexis Scale, the Measurement Instrument for Primary and Secondary Control Strategies, the Tennessee Self-Concept Scale, and the Bachman and O'Malley (1977) version of Rosenberg's Self-Esteem Scale, the authors examined 106 community-dwelling women between the ages of 20 and 65. The women were categorized into three groups: young adulthood (20-34 years), middle adulthood (35-49 years), and older adulthood (50-65 years). Results indicated that all groups were similarly dissatisfied with their bodies. The oldest group, however, had significantly more cognitive control than the other two groups and had significantly higher self-esteem and self-concept than the youngest group of women. Against expectations, there was no significant difference between groups on body importance. Finally, a significant negative relationship was found between body dissatisfaction and both self-concept and self-esteem. However, it was found that age moderated this relationship, such that the relationships between body dissatisfaction and both self-concept and self-esteem were weakest for the oldest group and strongest for the youngest group. The authors concluded that, as women age, their self-image is somewhat protected from their body's increasing departure from the youthful thin ideal through greater cognitive control. In other words, women who exhibit greater control adopt cognitive strategies such as lowering their expectations or reappraisals, thereby increasing their acceptance of otherwise socially undesirable and largely uncontrollable age-related body changes that, in turn, maintains their self-concept and self-esteem. Body Image 16 The finding of no significant differences among different age groups of women on body dissatisfaction is in accord with several previous findings. Although the finding that body importance remained stable across the three groups in Webster and Tiggemann's (2003) study is in agreement with research by Paxton and Phythian (1999), who found no relationship between age and appearance importance for females, it is counter to the findings of Pliner, Chaiken, and Flett (1990) and Tiggemann and Lynch (2001), who found that appearance importance decreased with age. There could be a variety of reasons for these mixed findings, including differences resulting from the different age boundaries used to group participants and the use of different measures of appearance importance were used in each study. It is possible that the measures did not measure the same construct equally well, or the construct of appearance importance may not be unidimensional. That is, it is possible that appearance importance may be composed of two or more factors and the various measures may tap different aspects of this. Thus, two basic fundamental problems are a lack of consistency in the measures used to assess the construct of appearance importance across age groups and a possible lack of invariance of the measures used, which makes it difficult to compare the results of these studies. Clearly more research is needed to increase our understanding of the relationship between age and appearance importance and to replicate these findings using a consistent and invariant measure of appearance importance. Given the rather extreme beauty ideals that are so prominent in society and valued in women (e.g., Novack, 1997; Wilcox, 1997), it seems surprising that, as women age and continually move further from these ideals, research on women's body dissatisfaction fails to find consistent age differences. Some research suggests that, although body dissatisfaction may not decrease, there are other psychological changes that are occurring in older women. These changes include decreased appearance anxiety, habitual body monitoring, and self-Body Image 17 objectification (Tiggemann & Lynch, 2001) and greater cognitive control (Webster & Tiggemann, 2003). So, as women age, they do physically move further from the ideals set by society, which should potentially serve to increase body dissatisfaction; however, it appears that these women may modify the way they respond and react to the changes in their bodies, which acts to negate any potential increases in body dissatisfaction. Thus, the result is a consistent level of body dissatisfaction across the life span. Gender Similarities and Differences in Body Image in Middle-Aged and Older Adults It is quite apparent from the research that women at any age are dissatisfied with their bodies. The findings of the 1972, 1985, and 1997 National U.S. surveys on body image indicate that men too are developing a more negative image of themselves over time (Cash et al., 1986; Garner, 1997). However, very few studies have focused on body image concerns related specifically to older men. Of those that have been done, a few have looked at the issues of body image in older women and older men in general (Oberg & Tornstam, 2001; Ziebland, Robertson, Jay, & Neil, 2002) and others specifically have examined gender differences that exist in perceptions of body image among older individuals (Gutpa & Schork, 1993; Janelli, 1993; Paxton, & Phythian, 1999; Pliner et al., 1990; Rozin & Fallon, 1988; Wilcox, 1997). A qualitative study by Ziebland et al. (2002) examined body image and weight change in middle-age. Seventy-two men and women between the ages of 35 and 55 were interviewed about their experiences of weight change. Participants also completed figure rating scales as part of the interview. A large majority of the sample had stated they had tried to lose weight. A common theme was also a belief that weight gain was inevitable in middle age. Responses to the figure rating scales indicated that both men and women preferred figures that were smaller than their current figures and the range of desirable figures was quite small. Oberg and Tornstam (2001) examined aspects of personal identity and how they Body Image 18 related to body image. Their sample consisted of 1250 Swedish men and women between the ages of 20 and 85. Participants completed a variety of measures including figure rating scales, estimations of real and ideal weights, and a question concerning rejuvenation through cosmetic surgery. It was discovered that keeping a youthful look becomes more important with age. On the figure rating scales, both men and women chose ideal figures that were smaller than their current figures. These figures increased slightly with age but were relatively narrow in range across the life span. Although not significantly correlated with age, another interesting finding was that 16% of women and 8% of men would consider having cosmetic surgery in order to obtain a younger appearance. There appears to be quite a parallel between men and women and their attitudes towards body image. Both sexes indicated that they would like to lose weight, and they were also similar in their preferences for an ideal body shape that is significantly smaller than their current shape. However, as mentioned previously, the finding that men prefer smaller ideals may be slightly misleading as the authors only took into consideration the average preference of men's ideal body shape. As has been found with younger men, there are a substantial number of men who would prefer an ideal body shape that is larger than their current ideal. It is currently unknown whether this same pattern exists in middle-aged and older men. As both men and women tend to believe that there is an inevitable weight gain in their middle and later years (Ziebland et al., 2002) it could be hypothesized that it is only young men who sometimes prefer a larger ideal body shape. Wilcox (1997) compared men and women's self-perceptions of physical attractiveness across the adult life span to test the theory of the double-standard of aging. This theory proposes that, with increasing age, women will view their physical appearance more negatively than men because women are constantly bombarded with messages that to be Body Image 19 attractive and they need to appear youthful, whereas men are given messages that they are "allowed" to appear their age and that aging may even enhance a man's attractiveness (Nowak, 1977; Sontag, 1979). Using the MBSRQ, the Index of Self-Esteem, and the Personal Attributes Questionnaire, Wilcox tested 144 men and women aged 20 to 80. Participants were equally distributed in each of six 10-year age brackets. Results did not support the theory of the double standard of aging. No significant differences were found between men and women on how they rated their appearance. There were also no significant differences found for age. Janelli (1993) investigated the concept of body image as it applies to older adults in a long-term-care setting, with a focus on gender differences. A total of 89 men and women between the ages of 60 and 98 completed the Body-Cathexis-Self-Cathexis Scale and the Draw-A-Person Technique. The Body-Cathexis portion of the former scale is composed of lists of body parts and functions. The Self-Cathexis portion of the scale contains items believed to represent different aspects of self such as life goals, memory and happiness. The Body-Cathexis scale revealed no significant differences between older men and older women in satisfaction with body parts. The Self-Cathexis scale revealed that older women were significantly more satisfied than the men with aspects of the self. Results for the Draw-A-Person (DAP) Technique revealed that the women's drawings were smaller in area, shorter, and less centered than were the men's, suggesting that older women have a poorer body image perception than older men. The author claims this provides support for the idea that older women may have internalized the self-fulfilling prophecy that old is not beautiful. However, I question the validity of a measure that makes claims about women's body image based on the size of a drawing. I fail to see the connection between the quality of a drawing and a person's perception of their body image. Even the author notes that the DAP technique has inherent difficulties in establishing reliability and validity, yet she still chose to use the measure and Body Image 20 draw conclusions from it. Pliner, Chaiken, and Flett (1990) investigated gender differences in eating, weight, global self-esteem, appearance self-esteem, and physical appearance over the life span using the Eating Attitudes Test, the Feelings of Social Inadequacy Scale, the Personal Attributes Questionnaire, and the authors' own measure on importance of weight and appearance. Their sample consisted of 334 females and 305 males in Canada grouped in ages of 10-13, 14-18, 19-24, 25-29, 30-39, 40-59, and 60 and over. Their results indicated that females expressed significantly greater concern and anxiety about eating, and reported significantly lower scores for appearance self-esteem. Significant age differences were found for both global and appearance self-esteem, with both types of self-esteem increasing linearly with age. There were both gender and age differences for importance of appearance, with scores being higher for females and decreasing with age. There were no significant interactions. These important findings extend the research on gender differences seen formerly only in adolescent and young adult populations to apply across the life span. The authors found that, even in older groups of men and women, gender differences on eating habits, general self-esteem, appearance self-esteem, and importance of appearance do not disappear. Gupta and Schork (1993) examined the relation between aging-related concerns and body image in a nonclinical sample of 173 Canadian men and women. Participants, grouped into four age categories ranging from under 30 to over 65, were tested using the Drive for Thinness and Body Dissatisfaction subscales of the Eating Disorder Inventory as well as a measure developed by the authors to assess aging-related concerns. Pertinent findings included a strong positive correlation between Drive for Thinness scores and concern about the effects of aging on appearance. Many participants believed that weight loss would result in more youthful looks and a better body image. Women were found to be more concerned about Body Image 21 aging-related changes than were men. However, women between the ages of 31 and 45 were found to be less concerned about maintaining younger looking skin compared to the other groups of women. The authors suggested that it may be during this time that women are more focused on career and families instead of their appearance. Rozin and Fallon (1988) compared attitudes toward weight and body image in two generations of men and women. Generational differences were examined using figure ratings, questions dealing with concern for weight, and weight and eating-related behaviors in a sample of 97 college-aged adults and their parents. Overall, both groups of women reported being more dissatisfied with their body image than were the men, and the older generation reported being more dissatisfied than the younger generation. This study is one of the few that has found significant differences between age groups for body dissatisfaction. A sex by generation interaction was also found indicating that the daughters, mothers, and fathers each had a substantial current-ideal disparity, whereas there was no disparity for the sons. For the sons, their current and ideal figures were almost the same. For each of the other groups, the ideal figure was much smaller than the current figure. The finding of no discrepancy for the young men is in accordance with the earlier findings of Fallon and Rozin (1985), but must be interpreted with caution, as the sign of the discrepancy score was not taken into account. However, the failure to find a similar pattern in the older group of men may suggest that most older men would like to lose weight and not gain weight. Paxton and Phythian (1999) examined gender differences in body image, self-esteem, and health status between middle and later adulthood. One hundred and fifty-nine Australian men and women, aged 40-79, were divided into middle and older adulthood based on the mean age of 56 years. Participants completed seven subscales of the MBSRQ, the three subscales of the Ben-Tovin Walker Body Attitudes Questionnaire (BAQ), the Rosenberg Self-Esteem Body Image 22 Scale, and the Self-Rated Health Status Scale. Pertinent results indicated that age for males correlated negatively with the appearance evaluation, appearance orientation, and fitness evaluation subscales of the MBSRQ and positively with the health orientation subscale of the MBSRQ. There were no significant correlations for females. In examining effects of gender and age, results indicated that females scored higher than males for the appearance orientation and health orientation subscales of the MBSRQ and for the disparagement, feeling fat, and salience of weight/shape subscales of the B A Q . Results also indicated that the middle-aged group scored higher on the appearance evaluation, appearance orientation, and health evaluation subscales of the MBSRQ, and lower on the disparagement subscale of the B A Q , when compared to the older group. A n interaction was also found for the fitness evaluation subscale of the MBSRQ and for Body Mass Index, with females scoring higher with age and males scoring lower with age. Although cohort effects cannot be ruled out, these results provide very different patterns of body image in males and females across these two age groups. These results also suggest that the body image of males may be more susceptible to the aging process as they are more likely to experience a decrease in feelings of attractiveness as age increases (Paxton & Phythian, 1999). Older males were also found to have a modest decline in the value placed on appearance, which is in accord with the findings of Pliner and colleagues (1990) who also found a decrease in appearance importance in older men. The failure to find a significant correlation between age and appearance orientation for females is consistent with the findings of Webster and Tiggemann (2003) and counter to the findings of Pliner et al. (1990) and Tiggeman and Lynch (2001) who found a decrease in appearance importance in older groups of women. As mentioned previously, these mixed findings may have resulted from the fact that all authors used different measures to assess appearance importance. Body Image 23 Taken together, the research on gender differences in body image among middle-aged and older adults has produced some mixed results. Some research has found no significant differences between men and women (Wilcox, 1997), yet other studies have found a range of differences between older men and women, including differences in body satisfaction (Janelli, 1993; Rozin & Fallon, 1988), concern about their appearance (Gutpa & Schork, 1993; Paxton & Phythian, 1999; Pliner et al., 1990), body attitudes, and investment in a healthy lifestyle (Paxton & Phythian, 1999). Gender differences, such as body dissatisfaction and concern about appearance, are consistent with research done with college samples whereby young women were found to have more negative body-image evaluations, stronger investments in their looks, and more frequent body-image dysphoria when compared to young men (Muth & Cash, 1997). In sum, although the limited research on body image in older individuals shows many similarities in perceptions of body image across the life span, there are certain aspects of body image that appear to distinguish older men and women from their younger counterparts, and from one another. On the one hand, both younger and older women tend to report similar levels of body dissatisfaction (Deeks & McCabe, 2001; Lewis & Cachelin, 2001) and tend to choose ideal body shape figures that are smaller than their current figure (Deeks & McCabe, 2001; Lewis & Cachelin, 2001; Oberg & Tornstam, 2001; Tiggemann & Lynch, 2001). On the other hand, younger and older women differ in that older women tend to be more negative about their appearance (Deeks & McCabe, 2001), show decreased appearance anxiety, habitual body monitoring, and self-obj edification (Tiggemann & Lynch, 2001), and demonstrate greater cognitive control (Webster & Tiggemann, 2003). A n issue that is prominent among both older men and older women is the believed inevitability of weight gain, demonstrated by the selection of progressively larger ideal body shape figures in figure rating scales. However, Body Image 24 these ideal body shapes are within a limited acceptable range for both genders (Deeks & McCabe, 2001; Lewis & Cachelin, 2001; Ziebland et al., 2002). Older women also differ from older men in that they tend to be more dissatisfied with their bodies and are, in general, more negative about their appearance. These findings suggest that are there are differences in body image among both men and women, whether it be age-related or cohort-related. However, a serious flaw in all these studies, and an issue that has not been addressed in this literature is whether or not the measures used are invariant in all age and gender groups. A l l too often, measures are used with the assumption that they are measuring the same concept(s) across groups. However, when a measure is used to make comparisons between two or more age groups, evidence is needed to show that the measure is functioning the same way in each age group. This evidence is accumulated by establishing equivalency across several aspects of the measure, such as construct measurement, construct interpretation, response format, and instructions. While the purpose of this thesis is not to look at all of these areas, ensuring that the construct is being measured equally well across groups is a logical initial step in providing evidence of measurement invariance and will be the focus of the present study. Measurement Invariance in Cross-Group Research Horn and McArdle (1992) have argued that, in any study involving the comparison of different groups, it is necessary to show measurement invariance before valid inferences and interpretations can be made. Measurement invariance refers to "whether or not, under different conditions of observing and studying phenomena, measurement operations yield measures of the same attribute" (Horn & McArdle, 1992, p. 117). Only through the demonstration of measurement invariance can a measure be deemed to be measuring the same attribute across groups. If there is no evidence of the presence or absence of measurement invariance, or i f Body Image 25 invariance is not obtained, any differences found between groups cannot be interpreted unambiguously. For example, age differences on a scale measuring appearance importance might be due to true differences between age groups on how important they rate their appearance, or they might be due to systematic biases in the way people of different ages respond to certain items, such that a response of "2" on an item means something different for younger individuals than it does for older individuals. Alternatively, findings of no differences between groups are also open to corresponding alternative interpretations. In short, as stated by Horn (1991), "without evidence of measurement invariance, the conclusions of a study must be weak" (p. 119). Evidence of measurement invariance is accumulated on an incremental basis. The weakest form of invariance, configural invariance, assesses whether the configuration of the salient and nonsalient factor loadings are equivalent across groups. In other words, the same items must load on the same factor(s) across all groups. In addition, those items that should not load on the factor should have loadings that are near zero across all groups, and the correlations between the factors (if more than one factor) must be substantially below 1.0. This last requirement is necessary to show that there is discriminant validity between the (sub) factors comprising the studied construct. Configural invariance is the minimum condition for factorial invariance. When configural invariance is met, it indicates that different groups associate the same subset of items with the same constructs. Steenkamp and Baumgartner (1998) argued that, although configural invariance supports that the same attribute is being measured in each group, it does not rule out the possibility that the items may not be measuring the attribute equally well in each of the groups and that respondents may not be interpreting the items in the same way. Thus, meaningful comparisons still may not be made across groups. Horn, McArdle, and Mason (1983) and Horn and McArdle (1992) have Body Image 26 questioned whether this assertion is too strict. They contended that, in all practicality, factor loadings will vary from sample to sample, and only the configuration of the zero and non-zero pattern coefficients realistically can be expected to remain invariant. Having the requirement that all items be interpreted in the same way by all respondents makes it nearly impossible to make any cross-group comparisons. Invariance is said to be achieved as long as variables unrelated to a factor in one sample are unrelated to that factor in all other samples. Thus, according to Horn et al. (1983), for all intents and purposes, when Configural invariance is met, one can be confident that the instrument is measuring the same construct across groups. While I believe that configural invariance is the most important step in the process of testing for measurement invariance, and that evidence of configural invariance is essential for ensuring that the construct is fundamentally similar across groups, it is only the first step. Having only configural invariance for a scale does allow one to use that scale in different groups to measure the construct of interest, and one can be confident that individuals are ascribing the same general meaning to the items and that the scale is generally measuring the same construct in each group. Nevertheless, the scale metrics may differ which results in the construct not being measured equally well in all groups. Thus, I feel configural invariance is not strong enough on its own to allow for comparisons across groups. Even though it is difficult to achieve higher levels of invariance, this does not mean that these higher levels are not necessary, especially i f the goal is to make accurate comparisons across groups. The next level of invariance is metric invariance. Metric invariance assesses whether there is equality of the unstandardized factor pattern weights (i.e., factor loadings) across groups. In other words, each group should have the same items load on the same factors and the factor loadings should be the same, or at least proportional to one another. This level of invariance provides a strong basis for inference that individuals from each of the groups Body Image 27 interpret and respond to the measure in a similar way (Horn & McArdle, 1992). That is, metric invariance provides evidence suggesting that the items are on equal scale intervals across age groups (Rock, Werts, & Flaugher, 1978). When metric invariance is met, the scale metrics are specified to be the same and thus the effects are, in principle, comparable. However, the invariance constraints on the scale metrics may not be justified and the estimated effects may be distorted (Baumgartner & Steenkamp, 2004). That is, the latent construct is assigned a scale by choosing a marker item and setting its loading to one. The intercept of the marker item is then set to zero, equating the mean of the latent construct to the mean of the marker variable. While this helps with identifying the model, it is not sufficient for meaningful comparisons. Thus, it is not advisable to make cross-group comparisons with only metric invariance. However, when metric invariance is met, one can examine structural relationships between the construct of interest and other constructs across groups (Steenkamp & Baumgartner, 1998). That is, one can examine and compare the strength of correlations between the construct of interest and other constructs across groups. In most research focusing on cross-group differences, it is important to conduct mean comparisons across groups. Thus, the third level of invariance, scalar invariance, assesses whether there is consistency between group differences in latent means and group differences in observed means. By examining the equality of the item intercepts, this level of invariance evaluates whether group differences in means are conveyed through common factors. Although testing for metric invariance assesses whether there are equal metrics or scale intervals across groups, there is still the possibility that the responses to items from one group can be systematically biased upward or downward (i.e., show additive bias) making comparisons of observed means across group uninterpretable (Meredith, 1993, 1995). Tests of scalar invariance circumvent this issue by removing this bias from the data. Meredith (1993) Body Image 28 and Steenkamp and Baumgartner (1998) insist that evidence of scalar invariance is necessary to make mean comparisons across groups. When scalar invariance is met, cross-group differences in the means of the observed items reflect differences in the means of the underlying construct(s). This level of invariance provides strong statistical evidence that the measure is invariant across groups and that valid comparisons across groups can be made. The highest levels of invariance testing include factor covariance invariance, factor variance invariance, and error variance invariance (Meredith, 1993; Steenkamp & Baumgartner, 1998). As the names imply, factor covariance invariance and factor variance invariance assess equality of the factor covariances and factor variances, respectively. If both are found to be invariant, it indicates that the correlations between the latent constructs are invariant across groups. Error variance invariance assesses equality of the amount of measurement error across groups. If this is found to be invariant, it implies that the items are equally reliable across groups. As scalar invariance is the highest level of invariance needed to conduct mean comparisons, these highest levels of invariance will not be further addressed in this paper (Steenkamp & Baumgartner, 1998; Yoo, 2002). For each level of invariance, one generally tests whether a given level of invariance is fully satisfied or not. Although recent attention has been given to the concept of partial measurement invariance, whereby only a subset of the scale items needs to meet the requirements of invariance in order for further invariance tests to be conducted (Byrne, Shavelson, & Muthen, 1989), concerns about the implications of partial invariance for measurement interpretation and the usefulness of a scale that is only partially invariant may limit its use. Given that the purpose of this paper is to assess whether three existing measures of body image can be used 'as is' to make comparisons across gender and age, tests for partial measurement invariance will not be made. Body Image 29 Present Study The majority of research on body image has been accumulated from adolescents and young adult college samples. However, with over 47% of Canada's population being 40 years of age or older (Statistics Canada, 2004), there is a great need for more body image research with older populations. The first step in targeting this population is to identify body image measures from which reliable and valid inferences can be made for adults in their middle and later years. Evidence of validity is provided on an incremental level and one source of validity evidence can be provided by assessing whether existing measures are invariant, or functioning equally well, among these older groups. If existing measures, developed primarily with younger samples, are found to demonstrate configural invariance in older samples, then these measures can be used to assess the same concepts in older samples. If measures are found to demonstrate metric invariance, then they can also be used to examine correlations with other constructs in older samples. If measures are found to exhibit scalar invariance, then they can also be used to make comparisons to younger samples. The body image measures chosen for the present study include the Multidimensional Body-Self Relations Questionnaire (MBSRQ), the Appearance Schemas Inventory-Revised (ASI-R), and the Body Image Quality of Life Inventory (BIQLI). The MBSRQ was selected for the present study because it is one of the most widely used measures of body image. It is also one of the more commonly used measures in the older adult body image literature mentioned previously, with only figure rating scales being used more often. No research, however, has specifically examined whether the subscales of this questionnaire are functioning equivalently across age and/or gender groups, and whether or not age or gender comparisons can be validly made. Unlike most body image measures, the standardization sample for the MBSRQ was selected to be representative of the U.S. Body Image 30 population in terms of gender and age; thus, it is one of the few measures not standardized with only college-aged students. However, Horn and McArdle (1992) point out that, just because a measure has been standardized with a varied sample, it does not mean that meaningful comparisons can be made across subsamples. Evidence of measurement invariance is still needed to ensure that valid group comparisons can be made. The ASI-R was included in the present study because of its focus on the importance of appearance in one's life. This may be of particular interest in the study of older men and women because of the contradictory research on appearance importance. As mentioned previously, several studies have examined the relationship between age and appearance importance and have found mixed results. Because of the lack of consistency in the measures used to assess appearance importance, it is difficult to interpret these findings. What is needed is a measure that can be consistently applied to further explore this issue. Before such a measure can be used, evidence is needed to show that this scale is measuring the same construct equally well in each group and that comparisons can be meaningfully conducted. Finally, the BIQLI was chosen for the present study because of its appeal for the study of individuals of all ages, and especially middle-aged and older adults. Body image quality of life refers to various life domains, such as interpersonal relationships, emotional states, diet and exercise, that are likely to be impacted by how one perceives one's body. As individuals age, there are multitudes of changes, both physical and psychological, that accompany the aging process that are likely to have an impact on body image quality of life. Thus, it would be of interest to investigate whether middle-aged and older adults differ significantly from their young adult counterparts in body image quality of life. In order for the MBSRQ, ASI-R, and BIQLI to be used to make comparisons across age or gender groups, it is necessary to demonstrate that these measures demonstrate scalar Body Image 31 invariance across each group. Therefore, the purpose of this study is to examine the configural, metric, and scalar invariance of each of these measures across three age groups of men and women. The age ranges in each group were chosen to roughly correspond to young adulthood (18-29)', middle-age (30-54), and older adulthood (55+). By examining the three levels of invariance, one can determine whether these measures are equivalent across these groups and can be used to make gender and age group comparisons. Several confirmatory factor analyses, performed at the multiple group level will provide the basis for evaluating these three scales' cross-group applicability. First, it will be ascertained whether the same number of factors are identified by the same observed variables in each of the different samples (i.e., young men, young women, middle-aged men, middle-aged women, older men, older women). This represents a necessary preliminary step for testing generalizability across samples and establishes the equivalency of factor patterns, or configural invariance. As mentioned in the literature, the M BSRQ has 10 factors, the ASI-R has two factors, and the BIQLI has one factor, as established in predominantly young adult samples. As the 10 factors of the MBSRQ do not add to a total score and often only some subscales are selected for use, this scale will be assessed by examining each of the subscales independently for measurement invariance. However, as two of the subscales (Fitness Evaluation and Self-Classified Weight) are composed of three items or less, tests of measurement invariance cannot be conducted for these subscales. Thus, these subscales will not be included in the analysis. To show configural invariance, one would expect to find the same factor pattern for each scale/subscale for each age/gender combination. Second, metric invariance or the invariance of factor loadings across samples will be investigated. Examination of factor loading invariance shows whether the 1 Young adults were categorized as 18 -29 because the majority of research in the body image field has focused on young adults in their late teens and early 20s. It was felt that this age grouping would make the group of young adults in the present study more comparable to previous research. Body Image 32 measures indicate the same factors to the same extent across samples. Finally, scalar invariance will be tested by examining the fit of the model, when the vector of item intercepts is made invariant across samples. Evidence of scalar invariance would indicate that there is consistency between the latent means and the observed means, and that cross-group comparisons can be validly made. If any of the scales fail to meet the minimum requirement of configural invariance, exploratory principal components analyses will be conducted separately for each group to determine the factor pattern. Body Image 33 Chapter Three Method Participants A total of 1344 participants were recruited for this study. Ten participants were immediately excluded from the data set for submitting questionnaires that had an excess of missing data for all three scales of interest (i.e., less than half of any scale completed). Participants were also excluded from the study i f they indicated they belonged to any of the following exclusionary criteria: were pregnant (n = 9), had been pregnant or given birth in the last year (n = 30), or had been clinically diagnosed within the last three years with anorexia (n = 10) or bulimia (n = 11). These participants were excluded because it was thought that these factors could have a large negative impact on an individual's perception of their body image and may bias the findings of the present study. For instance, those women who are pregnant or have recently given birth will have noticed changes in their appearance and body shape that would most likely result in a body image that is different from the usual image they had of themselves before they were pregnant. The one-year time period was chosen as it was felt this would give most women's bodies time to recover from the changes that occurred during and after pregnancy. For those women who identified themselves as having an eating disorder, it was felt that these women would have a distorted view of their body image and would not be representative of the general population, which is the focus of this study. Twelve participants were also excluded for not including their age and/or gender, as these two pieces of information were central to the analyses that were conducted. The final sample consisted of 1262 participants (422 men, 840 women). The men ranged in age from 18 to 98 (M= 39.7, SD = 19.1 years) and the women ranged in age from 18 to 89 (M= 39.4, SD = 18.7 years). See Table 1 for a summary of participant characteristics by gender and age group. Body Image 34 Table 1 Demographic Characteristics of Participants by Gender and Age Group (N = 1262) Young (18-29) Middle-Aged (30-54) Older (55+) Men Women Men Women Men Women Characteristic (n = 185) (n = 364) ( « = 1 3 1 ) (n = 267) (n = 106) (« = 209) Age M 22.3 22.8 42.3 40.6 67.2 67.0 SD 3.3 3.3 7.4 7.5 8.4 8.9 Ethnicity % White 50.8 63.7 77.9 86.9 84.9 92.3 % Non-White 48.6 36.3 22.1 13.1 15.1 6.7 Marital Status % Married 9.7 16.5 62.6 64.4 70.8 49.0 % Not Married 89.2 83.0 36.6 35.2 24.5 48.6 Education -% Grade 12 or Less 23.8 13.7 12.2 9.4 15.1 26.0 % Greater than Grade 12 76.2 86.0 87.8 90.6 84.9 72.6 Born in North America % Yes 70.8 83.8 70.2 81.6 68.9 67.0 %No 29.2 15.9 29.8 17.6 31.1 31.6 For the sample overall, 74.7% of participants identified themselves as White, 12.5% as East Asian, 5.5% as South Asian or Middle Eastern, 1.3% as Hispanic, 0.5% as Aboriginal or First Nations, and 5.2% as Other. Three participants did not indicate their ethnic background. Nearly 45% of the participants indicated that they were never married, 40.3% were married or Body Image 35 common-law, 4.3% were widowed, and 9.4% were divorced or separated. Sixteen participants did not indicate their marital status. Participants tended to be well educated with 2.5% having completed less than high school, 13.8% having completed high school, 42.3% having completed some college or university, 24.7% having completed a bachelor's degree, and 16.4% having completed a master's or PhD. Four participants did not indicate their education level. Seventy-six percent of the sample was born in North America. For the women, 31.9% had started or gone through menopause. Measures Multidimensional Body-Self Relations Questionnaire. The MBSRQ is a self-report inventory that assesses cognitive, behavioral and affective components of body image (Cash et al., 1986; Cash, 2000). Designed for adults and adolescents aged 15 and older, the 69-item MB SR Q consists of 10 subscales. Participants respond to all subscales using a 5-point Likert-type response format; however, a variety of anchor points are used (i.e., definitely agree to definitely disagree, never to very often, very underweight to very overweight, very dissatisfied to very satisfied). Scores for each of the subscales are calculated by taking the mean of its corresponding items. Thus, scores for each subscale range from 1 to 5. Appearance Evaluation (7 items) assesses feelings toward physical appearance, with higher scores reflecting greater satisfaction with appearance. Appearance Orientation (12 items) assesses investment in appearance, with higher scores indicating more importance and attention placed on looks and more engagement in grooming activities. Fitness Evaluation (3 items) assesses feelings towards physical fitness levels, with higher scores reflecting more physical dexterity. Fitness Orientation (13 items) assesses investment in fitness level, with higher scores reflecting more value placed on fitness and more involvement in fitness activities. Health Evaluation (6 items) assesses feelings Body Image 36 towards health, with higher scores indicating perceptions of a healthy body. Health Orientation (8 items) assesses investment in a healthy lifestyle, with higher scores reflecting a more "health conscious" lifestyle. Illness orientation (5 items) assesses reactivity to being or becoming i l l , with higher scores indicating a greater awareness of personal symptoms of physical illness and a greater likelihood to seek medical attention. The Body Areas Satisfaction Scale (9 items) assesses satisfaction with discrete aspects of appearance, with higher scores indicating contentment with more areas of one's body. The Overweight Preoccupation Scale (4 items) assesses a construct reflecting fat anxiety, weight vigilance, dieting, and eating restraint, with higher scores reflecting more of the construct. The Self-Classified Weight Scale (2 items) assesses perception and labeling of weight, from very underweight to very overweight (Cash, 2000) Appearance Schemas Inventory-Revised. The ASI-R (Cash, 2003; Cash, Melnyk, & Hrabosky, 2004) is a self-report measure of dysfunctional investment in appearance. This 20-item measure is composed of two subscales: Self-Evaluative Salience (12 items), which is the extent to which individuals' beliefs about their looks influence their personal or social worth and sense of self, and Motivational Salience (8 items), which is the extent to which individuals attend to their appearance and engage in appearance-management behaviors (Cash et al., 2004). Participants respond to each item on a 5-point Likert-type rating scale ranging from strongly disagree (1) to strongly agree (5). The Self-Evaluative Salience and Motivational Salience subscales are computed by taking the means of their respective items. A higher Self-Evaluative Salience score reflects a greater investment in physical appearance. A higher Motivational Salience score reflects a greater engagement in appearance-management behaviors. A Composite ASI-R score can also be calculated by taking the mean of all 20 items. A higher Composite score reflects greater Body Image 37 levels of overall dysfunctional schematic investment in appearance. Body Image Quality of Life Inventory. The BIQLI (Cash & Fleming, 2002) is a self-report measure of the impact of body image on a variety of life domains. It consists of 19 items measuring the domains of "life in general, emotional states, same and other sex relations, eating and exercise, grooming activities, sexual experiences, and family and work/school contexts" (Cash & Fleming, 2002, p. 458). For each question, participants rate on a 7-point bipolar scale how appearance affects that aspect of their life. Responses range from -3 (very negative effect) to 0 (no effect) to +3 (very positive effect). A composite score for the BIQLI is calculated by taking the mean of all 19 items. Possible scores can range from -3 to +3. A higher score reflects a more positive impact of body image on an individual's quality of life. Personal demographic form. A personal demographics form was included at the end of each questionnaire packet. The demographic form inquired about each participant's age, gender, level of education, marital status, ethnic/racial/cultural background, medical history, height and weight. This information was collected for the purpose of selecting and describing the sample. A copy of the personal demographic form may be found in Appendix A . As the MBSRQ, ASI-R, and BIQLI are copyrighted, copies of these measures are not provided. Individuals interested in purchasing these measures may do so at http://www.bodv-images.com. Procedure Data collection for this study occurred over a six-month period and was conducted in two forms: a web-based survey (n = 819) and a paper and pencil survey (n = 443). The content of the two surveys was identical. Initially, it was intended to collect the majority of the data Body Image 38 using the Internet, as a web-based survey allows for greater access to a larger and more varied sample than more traditional techniques. However, it was anticipated and soon confirmed that few men and women over the age of 55 were completing the web-survey, possibly as a result of not having Internet access, thus this age group was largely targeted with the paper and pencil format. Paper and pencil versions were also used to target college and university students in a classroom setting. Although there has been some concern about the validity of web-based surveys, Gosling, Vazire, Srivastava, and John (2004) have shown that the quality of data from Internet samples does not differ significantly from data collected from more traditional samples. In their research, they found that the characteristics of the participants were actually more diverse using Internet methods, differences in presentation formats did not appear to jeopardize the nature of the results, and the effects obtained using Internet methods were consistent with the effects from studies using traditional methods. Recruitment for the web survey took place in a number of ways: "snowball sampling" via emails, posters distributed throughout the community, and oral announcementsmade in psychology and human kinetics classrooms, community centres, senior citizens centres, and shopping malls. Individuals were provided with a link to the survey materials and interested individuals completed the survey at a time and location of their choice. A l l participants viewed an electronic informed consent form and gave their consent by completing and submitting the survey materials. A l l information was collected on a secure surver. Once data collection was complete, all survey materials were removed from the Internet. Recruitment for the paper and pencil version of the survey also took place through posters distributed throughout the community, and oral classroom and community announcements. An envelope containing the research materials was provided to interested individuals. In some cases (i.e., classroom setting), participants completed the survey right Body Image 39 away but, in most cases, participants took the materials home and completed the survey at a time and place of their choosing to maintain a similar research setting to those participants who chose to complete the study online. A n attempt was also made to refrain from answering questions about specific items in order to again maintain a similar research environment to those participants completing the survey online. A l l participants viewed a consent cover letter and gave their consent by completing and submitting the survey materials. Once the survey was complete, participants were instructed to return the materials to the envelope and seal the envelope. Participants were asked to return the package to a place set by the researcher or research assistant by a specified time. A l l participants received the questionnaires in the same order: Multidimensional Body-Self Relations Questionnaire, Appearance Schemas Inventory-Revised, Body Image Quality of Life Inventory, a Self-Efficacy Scale, a Subjective Age Identity Scale, and a personal demographics sheet. Completion of the survey took approximately 30 minutes. The Self-Efficacy Scale and the Subjective Age Identity Scale were not used in the present study. Model Evaluation A l l tests of measurement invariance were investigated with multi-group confirmatory factor analysis (MGCFA) using LISREL (Joreskog & Sorbom, 1993). The maximum likelihood (ML) estimation method with a Pearson product moment (ppm) covariance matrix was used to analyze the data. A covariance matrix was chosen over a correlation matrix because, as Cudeck (1989) noted, the analysis of correlation matrices may result in the modification of the model being analyzed, incorrect chi-square and other goodness-of-fit measures, and incorrect standard errors. Although a polychoric matrix is designed to deal with ordinal data, a ppm covariance matrix was chosen because a polychoric matrix has been found, when used with the M L method, to produce a higher frequency of nonpositive-definite Body Image 40 matrices, require larger sample sizes (Olsson, 1979) and produce overestimated test statistics and standard errors (Babakus, Ferguson, & Joreskog, 1987; Dolan 1994; Rigdon & Ferguson, 1991). The M L approach was chosen over other estimation methods for several reasons. First, although the data were found to be moderately non-normal, the M L approach has been shown to be robust under conditions of univariate skewness values less than 2.0 and kurtosis values less than 6.0 (Benson & Fleishman, 1994; Curran, West, & Finch, 1996; Muthen & Kaplan, 1985, 1992). A l l items for all groups met these criteria, except for two items on the M BSRQ for the group of young men. Second, although estimation methods such as weighted least squares (WLS) or categorical variable modeling (CVM) were developed to deal with ordinal data, the relatively small group sample sizes in the present study prevented the use of these methods, which require sample sizes of 2,000-5,000 participants per group (Browne, 1984). Third, in a simulation study by Dolan (1994), the M L method was found to perform quite well when the number of scale points was equal to seven, and moderately well the number of scale points was equal to five. In the present study, both the M BSRQ and the ASI-R have five scale points and the BIQLI has seven scale points. For each scale, up to a total of six models were tested. The first model that was tested in all cases, herein called the full model, was the model that tested the three age groups (young adult, middle-aged adult, older adult) by two gender groups (males, females). If the full model did not meet invariance requirements, then five age and gender subgroups were tested. The female model tested each female group across the three age groups. The male model tested each male group across the three age groups. The young adult model tested the males and females at the youngest age group. The middle-aged adult model tested the males and females at the middle age group. The older adult model tested the males and females at the oldest age group. Body Image 41 Overall model fit was assessed by examining a number of goodness-of-fit indices in addition to the chi-square test because a statistically significant chi-square can result even though there are only minor differences between the groups' factor patterns due to this statistic's sensitivity to sample size (Bollen & Long, 1993; Hu & Bentler, 1993). Steenkamp and Baumgartner (1998) recommend using the following four alternative fit indices: the root mean square error of approximation (RMSEA; Browne & Cudek, 1993; Steiger, 1990), the Comparative Fit Index (CFI; Bentler, 1990), the non-normed fit index (NNFI, also called the Tucker-Lewis Index; Bentler & Bonnet, 1980), and the consistent Akaike information criteria (CAIC; Bozdogan, 1987). The R M S E A was used because it has been found to be sensitive to missspecified factor loadings, with values close to or lower than .06 indicating good fit (Hu & Bentler, 1998, 1999). However, Browne and Cudeck (1993) advise that values between .05 and .08 indicate acceptable fit and that values between .08 and .10 indicate mediocre fit. The CFI was used because it has been found to be sensitive to misspecified factor pattern coefficients (Hu & Bentler, 1998, 1999). Hu and Bentler (1998, 1999) recommend values close to .95 or greater to indicate good fit, although it has generally been accepted that values greater than .90 indicate acceptable fit (Vandenberg & Lance, 2000). The NNFI was used because it is both sensitive to model misspecification and it appropriately penalizes model complexity (i.e., gives preference to simpler models). Hu and Bentler (1998, 1999) recommend values close to .95 or greater to indicate good fit; however, Vandenberg and Lance (2000) recommend that values of .90 can be used as a lower bound in indicating acceptable fit. Finally, the CAIC was used because it has been shown to be particularly useful for purposes of model comparisons because it takes model parsimony into account (Diamantopoulos & Siguaw, 2000). A model CAIC value that is less than both the independent and saturated CAIC values is indicative of good model fit. Overall model fit has a role only in appraising the test Body Image 42 for configural invariance because the interpretation of these values for further levels of invariance testing has not been determined (B. D. Zumbo, personal communication, June 1, 2005). However, for completeness of the tables, the results of these goodness-of-fit indices will be presented for each level of invariance testing. For testing metric and scalar invariance, all models were placed in a hierarchical sequence of nested models so that systematic comparison tests could be conducted (Joreskog, 1971). Although the degree of invariance across nested models is most frequently assessed by chi-square difference tests, researchers have shown that differences in chi-square values are dependent on sample size (Brannick, 1995; Kelloway, 1995). Thus, Cheung and Rensvold (2002) also recommend using change in CFI to assess differences between the models, with values less than or equal to -.01 indicating model invariance. A critical value of less than .01 was used for the chi-square difference test. In cases where there was disagreement between the conclusions of the chi-square difference test and CFI difference test, the CFI difference test was given more weight. Figure 1 outlines the steps that were used to test for measurement invariance. A l l analyses were conducted separately for each measure. The first step was to test for configural invariance. Configural invariance indicates that all groups associate the same subsets of items with the same constructs. This is tested by examining i f the items of each of the scales exhibit significant nonzero loadings on salient factors and zero loadings on non-salient factors. If this model is found to fit well, it can serve as a baseline model for comparisons with more restricted models. A poorly fitting model at this step indicates that the groups conceptualize the latent construct differently. In the present study, i f configural invariance was not supported for the full model, configural invariance testing was done for the five subgroups (i.e., female model, male model, young adult model, middle-aged adult model, older adult model). For Configural for invariance for all subsets? no Configural invariance for subset of groups? no yes yes yes Compare means of all groups yes yes yes Compare means for invariant groups no no Figure 1. Procedure for assessing measurement invariance. Groups not comparable Groups not comparable Groups not comparable a o C L Body Image 44 those subgroups that met the hypothesis of configural invariance, further tests of metric invariance were conducted. For those subgroups that did not meet the hypothesis of configural invariance, principal components analyses (PCAs) were conducted on the individual groups to explore why configural invariance was not met. These PCAs were conducted because each of the scales used in the present study have a predefined number of factors and if configural invariance does not hold for any of the scales, this suggests that the factor structure specified may not be correct. Thus, the PCAs were conducted to examine whether a different factor pattern emerged in the data. The second test of invariance that was examined was metric invariance. Whereas configural invariance establishes that the groups relate the same items with the same factors, metric invariance tests the similarity of the strength of these relationships. This is tested by constraining the matrix of factor loadings to be invariant across groups and evaluating model fit. If this model is found to fit well, it implies that the groups perceive the latent construct in the same manner and that the indicators are on equal scale intervals across groups (Rock et al., 1978). A poorly fitting model at this stage can indicate that the latent construct is poorly operationalized or that cross group differences exist in how the latent construct is conceptualized (Cheung & Rensvold, 2000). In the present study, if metric invariance was not supported for the, full model, metric invariance testing was conducted for the five subgroups (i.e., female model, male model, young adult model, middle-aged adult model, older adult model). For those subgroups that met the hypothesis of metric invariance, further tests of scalar invariance were conducted. If metric invariance was not found for any subgroup, testing was stopped. The final test of invariance that was examined was scalar invariance. Scalar invariance indicates that individuals who have the same value on the latent construct would obtain the Body Image 45 same value on the observed variable, regardless of their group membership. This level of invariance is tested by constraining the vector of item intercepts across groups and examining model fit. If this model is found to fit well, then mean differences of observed scores can be compared, and such differences can be considered reflections of true differences between the groups on the latent trait. A poorly fitting model indicates that differential bias exists that is causing the groups to respond differently to the items. In the present study, i f scalar invariance was not supported for the full model, scalar invariance testing was conducted for the five subgroups (i.e., female model, male model, young adult model, middle-aged adult model, older adult model). For those subgroups that met the hypothesis of scalar invariance, mean score comparisons were conducted on that scale for those groups. If scalar invariance was not found for any subgroup, testing was stopped. Body Image 46 Chapter Four Results Missing and Ambiguous Data A fair amount of missing data was observed in this data set. Missing data could have resulted from one of three sources. First, in the questionnaire cover sheet participants were informed that they had the right to refuse to answer any specific question and so participants may have chosen not to respond to specific items or scales. Second, as participants were not monitored during the data collection process and questionnaires were not reviewed by the researcher or research assistants before being submitted, items and/or pages may have been mistakenly overlooked by the participant and omitted by accident. Third, in some cases, participants were not clear in their response to an item and either completed the item incorrectly or indicated more than one response for the item. For all three scales, i f a participant indicated their response as two adjoining options (i.e., 3-4), items were randomly scored, by the toss of a coin, as the lower or the higher value. For all other items falling under the third reason for missing data, the items could not be scored and were left as missing. Neither the manuals for the M BSRQ (Cash, 2000) or ASI-R (Cash, 2003) nor the current literature provide instructions for how to deal with missing responses for these scales. Therefore, I had to decide how I would deal with missing data responses for each body image scale prior to conducting analyses on each measure. It was arbitrarily decided to exclude those participants who had missed more than 20% of the scale, either from not finishing the scale, from missing a page, or from missing too many items. For the remaining participants, scores were imputed for the missing data, with the exception of BIQLI items 6, 11, and 12, which will be detailed below. Schafer and Olsen (1998) note that simple mean substitution for missing responses would seriously diminish relationships among variables; so missing data Body Image 47 were imputed by means of the expectation maximization method (EM method). The E M method estimates a mean vector and a covariance matrix on incomplete data through maximum likelihood estimation. This information then is used to calculate multiple regression equations to predict the values of the missing points, using all the variables that are present for each case. Since the E M method makes use of all available data, it is one of the most effective methods to impute missing values (Hox, 1999). Multidimensional Body-Self Relations Questionnaire. Of the total sample, 16 (4.0%) of the men (2.7% = young; 3.8% = middle-aged; 6.6% = older) and 47 (5.8%) of the women (2.7% = young; 4.1% = middle-aged; 13.4% = older) had missing responses on the MBSRQ. One participant completed less than half of the questionnaire and 5 participants missed a page for this scale. Thus, two (0.5%) of the men (1.1% = young, 0% = middle-aged, 0% = older) and four (0.5%) of the women (0% = young, 0% = middle-aged, 1.9% = older) had to be excluded from the analyses due to too many missing responses. Missing data was imputed for a total of 57 participants. Three participants (1 male, 2 females) each had nine missing responses, one female participant had five missing responses, two participants (1 male, 1 female) each had four missing responses, one male participant had three missing responses, eight participants (2 male, 6 female) each had two missing responses, and 42 participants (9 males, 33 females) each had one missing response. The items that were missed for this scale appeared to be somewhat randomly distributed throughout the scale with a total of 47 out of 69 items having at least one missing response and no single item having more than five missing responses. Appearance Schemas Inventory-Revised. Of the total sample, eight (1.9%) of the men (1.6% = young; 2.3% = middle-aged; 1.9% = older) and 14 (1.7%) of the women (0.8% -young; 2.2% = middle-aged; 2.4% = older) had missing responses on the ASI-R. Six Body Image 48 participants skipped the entire scale, one participant did not complete the scale, and four participants missed a page for this scale. Thus, two (0.5%) of the men (0.5% = young, 0.8% = middle-aged, 0% = older) and nine (1.1%) of the women (0% = young, 1.9% = middle-aged, 1.9% = older) were excluded from the analyses due to too many missing responses. Missing ' data was imputed for a total of 12 participants. One male participant had five missing responses, one female participant had two missing responses, and 10 participants (5 men, 5 women) each had one missing response. The items that were missed for this scale appeared to be somewhat randomly distributed with a total of 10 out of 20 items having at least one missing response, and no single item having more than four missing responses. Body Image Quality of Life Inventory. Of the total sample, 17 (4.0%) of the men (2.7% = young; 2.3% = middle-aged; 8.5% = older) and 61 (7.3%) of the women (1.9% = young; 3.0% = middle-aged; 22.8% = older) had missing responses on the BIQLI. Five participants (3 women, 2 men) skipped the entire scale, one female participant missed a page, and three female participants completed less than half the scale. Thus, two (0.5%) of the men (0.5% = young, 0% = middle-aged, 0.9% = older) and seven (0.8%) of the women (0% = young, 0% = middle-aged, 3.3% = older) were excluded from the analyses due to too many missing responses. Missing data for item 6, "My experiences at work or school", item 11, "My feelings of acceptability as a sexual partner", and item 12, "My enjoyment of my sex life" were particularly pronounced for the oldest group of women, with 9.6%, 9.1%, and 12.4% of the older women having missing data on these items, respectively. These three items accounted for 68% of the missing responses among the oldest group of women (i.e., 65 out of 95 missing responses), and also accounted for half of the missing responses among the oldest group of men (i.e., 6 out of 12 missing responses). Missing data for item 12 were also more pronounced Body Image 49 for women in all age groups compared to the men, with 3.8% of women and only 0.7% of men having a missing response for this item. As it was felt that it was not necessarily applicable to impute values for these three items, two separate analyses were conducted for this scale with (a) items 6, 11, and 12 removed from the scale, and (b) all items remaining in the scale, but those participants (6 men, 39 women) who had missing responses for at least one of items 6, 11, or 12 removed from the data set. These separate analyses were conducted because it is highly likely that the reason these items had missing responses was because they did not apply to the participant. For instance, an older person who has retired would be unable to answer a question that asks how their body image impacts their experiences at work or school. Therefore, it did not make sense to impute a value for these items because the number that would be imputed would really be meaningless as the item is not relevant to the participant. For the remaining items, missing data were imputed for a total of 24 participants. One female participant had five missing responses, 7 participants (2 men, 5 women) each had two missing responses, and 16 participants (7 men, 9 women) each had one missing response. Assumptions Before examining the factor structure of the responses for each group, the data were examined for normality using PRELIS 2.53 (Joreskog & Sorbom, 1993). To satisfy the assumption of normality, Kline (1998) suggested absolute values less than 3.0 for skewness and 8.0 for kurtosis. Bentler (1998) also suggests that multivariate kurtosis values should be less than 3.0. Table 2 presents the values for skewness, kurtosis, and multivariate kurtosis of the three body image scales. These values indicated that the responses were distributed within acceptable limits for all items for all three scales. Body Image 50 Table 2 Absolute Skewness, Kurtosis, and Multivariate Kurtosis Values for the MBSRQ, ASI-R, and BIQLI Group Skewness Kurtosis Multivariate Kurtosis MBSRQ Young Men Young Women Middle-Aged Men Middle-Aged Women Older Men Older Women ASI-R Young Men Young Women Middle-Aged Men Middle-Aged Women Older Men Older Women BIQLI Young Men Young Women Middle-Aged Men Middle-Aged Women Older Men Older Women 0.02-2.72 0.02-1.41 0.01-1.33 0.02-1.40 0.01-1.50 0.00-1.72 0.01-0.85 0.03-0.91 0.02-0.64 0.03-0.84 0.02-0.59 0.11-1.06 0.01-1.13 0.00-1.02 0.07-0.95 0.02-0.60 0.03-0.90 0.11-0.99 0.05-6.98 0.01-3.10 0.03-1.94 0.01-2.62 0.05-2.33 0.01-4.30 0.00-1.25 0.07-1.19 0.18-1.12 0.08-1.10 0.01-1.06 0.01-1.16 0.11-1.53 0.12-1.04 0.01-0.68 0.20-1.17 0.01-1.43 0.03-1.48 1.04 1.05 1.02 1.05 1.00 1.05 1.07 1.07 1.08 1.08 1.12 1.09 1.29 1.33 1.38 1.39 1.23 1.41 Body Image 51 Reduced BIQLI (minus items 6, 11, 12) Young Men 0.01-1.13 0.10-1.57 1.30 Young Women 0.01-1.00 0.11-1.03 1.33 Middle-Aged Men 0.07-0.95 0.01-0.68 1.38 Middle-Aged Women 0.09-0.61 0.24-1.16 1.38 Older Men 0.06-0.88 0.01-1.40 1.23 Older Women 0.20-0.99 0.01-1.46 1.52 The data for each gender and age group were also examined separately for the presence of multivariate outliers. For the MBSRQ, using Tabachnick and FidelPs (2001) recommendation of a .001 probability and 69 degrees of freedom, Mahalanobis' distance values over 111.05 indicated the presence of multivariate outliers. A total of 24 outliers were identified for this scale (young men = 3, young women = 8, middle-aged women = 8, older women = 5). For the ASI-R, values over 45.31 indicated the presence of multivariate outliers. A total of 20 outliers were identified for this scale (young men = 2, middle-aged men = 3, older men = 2, young women = 5, middle-aged women = 4, older women = 4). For the BIQLI with all 19 items, values over 43.82 indicated the presence of multivariate outliers. A total of 59 outliers were identified for this scale (young men = 10, middle-aged men = 8, older men = 4, young women =16, middle-aged women = 10, older women =11). For the reduced BIQLI, values over 39.25 indicated the presence of multivariate outliers. A total of 54 outliers were identified for this scale (young men = 8, middle-aged men = 6, older men = 2, young women = 15, middle-aged women =10, older women =13). Analyses run with, and without, the outliers did not differ in the conclusions drawn, thus the outliers were kept in the sample. Body Image 52 Invariance of the Multidimensional Body-Self Relations Questionnaire The MBSRQ consists of 10 separate subscales that can be used jointly or independently. As these subscales do not add up to a total score, and the full scale is not always used within a single study, all measurement invariance tests were conducted separately for each subscale. However, because the Fitness Evaluation and the Self-Classified Weight subscales have only three and two items, respectively, invariance testing could not be conducted on these subscales. Thus, only eight of the 10 subscales were investigated. Each subscale was assumed to be a unidimensional scale. Appearance Evaluation. Tables 3 and 4 summarize the results of the measurement invariance tests for the Appearance Evaluation subscale of the MBSRQ. The first test of invariance across the six age and gender groups was for configural invariance. Although the significant chi-square and R M S E A value indicated a poor fit of the model, the remaining three fit statistics (i.e., NNFI, CFI, CAIC) indicated an acceptable fit of the model (see Table 3). Given these mixed results, I based my decision for configural invariance on the fact that the majority of the fit statistics supported the fit of this model, and concluded that, although this model did not provide a great fit to the data, it did provide an acceptable fit and further tests of measurement invariance were warranted. Thus, it was concluded that the Appearance Evaluation subscale showed the same factor configuration across the six age and gender groups. As configural invariance was supported for the full model, the next step in the analysis was to test for metric invariance for the full model. The increase in chi-square from the indicated that the hypothesis of metric invariance was not tenable (see Table 4). Thus, metric invariance was not established for the full model. This finding indicates that the strength of the relationship between each item and factor was significantly different between the groups. Table 3 Goodness-of-Fit Indices for the MBSRQ - Appearance Evaluation Model I2 df R M S E A NNFI CFI Model CAIC Independence CAIC Saturated CAIC Configural - full model 454.28** 84 .15 .91 .94 1163.143 6989.35 1366.53 Metric - full model 630.75** 119 .15 .92 .92 1039.29 a 6989.35 1366.53 Female model 296.29** 56 .13 .95 .95 533.10a 5359.36 649.10 Male model 290.46** 56 .18 .82 .84 508.10a 1650.44 591.18 Young adult model 225.70** 35 .15 .93 .94 404.92a 3577.22 409.05 Middle-aged adult model 226.70** 35 .16 .89 .91 366.87a 2181.94 391.24 Older adult model 138.61** 35 .13 .89 .91 272.58 a 1234.81 377.07 Scalar - Female model 391.56** 66 .14 .94 .94 735.183 5454.40 649.20 Young adult model 316.91** 40 .18 .91 .92 598.65 3432.72 409.05 Note. a model CAIC < than both independence and saturated CAICs *p<.Q\. **/7<.001. Table 4 Tests of Measurement Invariance for the MBSRQ - Appearance Evaluation Model A x ' Adf ACFI Invariance? Configural - full model - - - yes Metric - full model 176.47** 35 -.02 no Female model 14.71 14 0 yes Male model 117.76** 14 -.06 no Young adult model 11.82 7 0 yes Middle-aged adult model 98.22** 7 -.04 no Older adult model 26.68** 7 -.02 no Scalar - Female model 109.98** 24 -.01 yes Young adult model 103.03** 12 -.02 no Note. Change scores are not calculated for the configural model because this is the baseline model. *p<.Q\. **p<.001. Body Image 55 As metric invariance was not supported for the full model, the next step was to assess whether metric invariance was established for any of the five age and gender subgroups. As summarized in Table 4, metric invariance was not met for the male model (i.e., young, middle-aged, and older men), middle-aged adult model (i.e., men and women within this age group), or older adult model (i.e., men and women within this age group), suggesting that the construct of appearance evaluation was not perceived in the same manner and that the items were not on equal scale intervals across the groups within each model. Thus, observed mean scores should not be compared on this subscale for these groups. Metric invariance was shown for the female model (i.e., young, middle-aged, and older women) and for the young adult model (i.e., men and women within this age group), however, indicating that tests of scalar invariance could be conducted for these two subgroups. The final measurement invariance test conducted was for scalar invariance for the female and young adult models. For the female model, the increase in chi-square from the configural invariance model was significant (A % 2 - 109.98,/? < .001), which indicated that the hypothesis of scalar invariance was not tenable. However, ACFI = -.01, which indicated that the hypothesis of scalar invariance was tenable. Given these contradictory findings, and giving more weight to ACFI because differences in chi-square values have been found to be dependent on sample size (Brannick, 1995; Kelloway, 1995), I concluded that scalar invariance was shown for the female model. This suggests that women responded to the items in the same way and that comparisons of observed scores on this subscale may be examined and interpreted across age groups. For the young adult model, the increase in chi-square from the configural invariance model was significant (A % 2 = 103.03,-/? < .001) and ACFI = -.02, which indicated that the hypothesis of scalar invariance was not tenable. This suggests that measurement bias exists Body Image 56 that caused the young men and women to respond differently to the items and that differences on observed scores for these groups should not be compared on this subscale. Appearance Orientation. Tables 5 and 6 summarize the results of the measurement invariance tests for the Appearance Orientation subscale of the MBSRQ. The first test of invariance across the six age and gender groups was for configural invariance. Although the chi-square was significant, all other fit indices (i.e., R M S E A , NNFI, CFI, CAIC) indicated an acceptable fit of the model (see Table 5). Thus, it was concluded that the Appearance Orientation subscale showed the same factor configuration across the six age and gender groups, and that further tests of measurement invariance were warranted. As configural invariance was supported for the full model, the next step in the analysis was to test for metric invariance for the full model. The increase in chi-square from the configural invariance model was significant (A % 2 = 180.18,/? < .001), which indicated that the hypothesis of metric invariance was not tenable. However, ACFI = -.01, which indicated that the hypothesis of metric invariance was tenable. Given these contradictory findings, and giving more weight to ACFI, I concluded that metric invariance was established for the full model. Thus, the strength of the relationship between each item and factor was not statistically different between the groups. As metric invariance was supported for the full model, the next step in the analysis was to test for scalar invariance for the full model. The increase in chi-square from the configural invariance model was significant (A % = 722.69, p< .001) and ACFI = -.05, which indicated that the hypothesis of scalar invariance was not tenable. Thus, scalar invariance was not established for the full model. This finding suggests that measurement bias exists that caused the groups to respond differently to the items. Table 5 Goodness-of-Fit Indices for the MBSRQ - Appearance Orientation Model i2 df R M S E A NNFI CFI Model CAIC Independence CAIC Saturated CAIC Configural - full model 819.51** 324 .09 .94 .95 2022.71s 11153.07 3806.76 Metric - full model 999.69** 384 .09 .94 .94 1727.67a 11153.07 3806.76 Scalar - full model 1542.20** 434 .12 .90 .89 2547.76a 11159.06 3807.13 Female model 835.83** 206 .11 .91 .91 1390.26a 7515.09 1808.50 Male model 523.94** 206 .11 .90 .90 985.90a 3589.85 1646.86 Young adult model 474.10** 130 .10 .91 .91 880.50a 4073.33 1139.49 Middle-aged adult model 352.30** 130 .09 .94 .91 701.13a 3793.63 1089.89 Older adult model 312.86** 130 .10 .94 .94 686.91 a 3211.06 1050.91 Note. a model CAIC < than both independence and saturated CAICs *p<M. **/?<.001. Table 6 Tests of Measurement Invariance for the MBSRQ - Appearance Orientation Model A x ' Adf ACFI Invariance? Configural - full model - - - yes Metric - full model 180.18* 60 -.01 yes Scalar - full model 722.69** 110 -.06 no Female model 296.62** 44 -.04 no Male model 242.84** 44 -.06 no Young adult model 137.74** 22 -.03 no Middle-aged adult model 109.27** 22 -.01 yes Older adult model 71.95** 22 -.02 no Note. Change scores are not calculated for the configural model because this is the baseline model. */?<.01. **/?<.001. Body Image 59 As scalar invariance was not supported for the full model, the final step was to assess whether scalar invariance was established for any of the five age and gender subgroups. As summarized in Table 6, scalar invariance was not met for the female, male, young adult, or older adult models, which indicated that measurement bias cannot be ruled out when comparing across groups. Thus, observed mean scores should not be compared on this subscale for these groups. Scalar invariance was only supported for the middle-aged adult model, which indicated that middle-aged men and women responded to the items in the same way and that comparisons of observed scores on this subscale may be examined and interpreted across gender for this age group. Fitness Orientation. Tables 7 and 8 summarize the results of the measurement invariance tests for the Fitness Orientation subscale of the MBSRQ. The first test of invariance across the six age and gender groups was for configural invariance. Although the significant chi-square and R M S E A indicated a poor fit of the model, the remaining three fit statistics (i.e., NNFI, CFI, CAIC) indicated an acceptable fit of the model and thus I concluded that, although this model did not provide a great fit to the data, it did provide an acceptable fit and further tests of measurement invariance were warranted (see Table 7). Thus, it was concluded that the Fitness Orientation subscale showed the same factor configuration across the six age and gender groups. As configural invariance was supported for the full model, the next step in the analysis was to test for metric invariance for the full model. The increase in chi-square from the configural invariance model was significant (A % 2 = 207.43,/? < .001), which indicated that the hypothesis of metric invariance was not tenable. However, ACFI = 0, which indicated that the hypothesis of metric invariance was tenable. Following my arguments in previous analyses, I concluded that metric invariance was shown for the full model. This finding indicates that the Table 7 Goodness-of-Fit Indices for the MBSRQ - Fitness Orientation Model x 2 df R M S E A NNFI CFI Model CAIC Independence CAIC Saturated CAIC Configural - full model 1372.06** 395 .11 .93 .94 1625.76a 17662.22 4441.22 Metric - full model 1534.49** 460 .11 .93 .94 2256.94a 17662.22 4441.22 Scalar - full model 1866.08** 510 .12 .92 .92 2945.70 a 17689.19 4441.65 Female model 1112.47** 243 .12 .93 .93 1768.37a 12701.51 2109.92 Male model 638.72** 243 .10 .91 .91 1096.63a 4929.03 1921.34 Young adult model 548.38** 154 .10 .95 .95 945.77a 8035.81 1329.41 Middle-aged adult model 515.83** 154 .11 .94 .94 921.14a 6065.63 1271.53 Older adult model 483.04** 154 .12 .89 .90 874.76a 3499.91 1226.06 Note. a model CAIC < than both independence and saturated CAICs */?<.01. **/?<.001. Table 8 Tests of Measurement Invariance for the MBSRQ - Fitness Orientation Model Adf ACFI Invariance? Configural - full model - - - yes Metric - full model 207.43** 65 0 yes Scalar - full model 493.62** 115 -.02 no Female model 285.39** 46 -.02 no Male model 95.96** 46 -.01 yes Young adult model 70.31** 23 0 yes Middle-aged adult model 63.69** 23 0 yes Older adult model 40.07 23 0 yes Note. Change scores are not calculated for the configural model because this is the baseline model. */?<.01. **/?<.001. Body Image 62 strength of the relationship between each item and factor was not statistically different between the groups. As metric invariance was supported for the full model, the next step in the analysis was to test for scalar invariance for the full model. The increase in chi-square from the configural invariance model was significant (A % 2 = 493.62,p < .001) and ACFI = -.02, which indicated that the hypothesis of scalar invariance was not tenable. Thus, scalar invariance was not met for the full model. This finding suggests that measurement bias exists that caused the groups to respond differently to the items. As scalar invariance was not supported for the full model, the final step was to assess whether scalar invariance was established for any of the five age and gender subgroups. These results are summarized in Table 8. Scalar invariance was not shown for the female model, which indicated that measurement bias cannot be ruled out when comparing women across age groups on this subscale. Thus, observed mean scores for women should not be compared on this subscale. Scalar invariance was demonstrated for the male, young adult, middle-aged adult, and older adult models, indicating that these groups responded to the items in the same way and that comparisons of observed scores on this subscale may be conducted and ( interpreted across gender for each age group and across age groups for men. Health Evaluation. Tables 9 and 10 summarize the results of the measurement invariance tests for the Health Evaluation subscale of the MBSRQ. Given that three of the five fit statistics (i.e., % , R M S E A , NNFI) do not support the fit of this model, I concluded that configural invariance was not supported for the full model (see Table 9). Thus, it was concluded that the Health Evaluation subscale did not show the same factor configuration across the six age and gender groups. Table 9 Goodness-of-Fit Indices for the MBSRQ - Health Evaluation Model Independence Saturated Model I2 df R M S E A NNFI CFI CAIC CAIC CAIC Configural - full model 270.63** 54 .14 .89 .93 865.18a 3654.48 1024.90 Female model 203.81** 27 .16 .88 .93 490.28 2686.18 486.83 Male model 66.82** 27 .10 .91 .95 320.80a 941.25 443.39 Young adult model 114.22** 18 .14 .88 .93 294.55a 1489.30 306.79 Middle-aged adult model 97.83** 18 .15 .86 .92 271.16a 1042.39 293.43 Older adult model 58.58** 18 .12 .93 .96 218.403 1082.26 282.80 Metric - Male model 101.69** 39 .10 .91 .92 266.35a 941.25 443.39 Older adult model 79.75** 24 .12 .93 .94 194.793 1085.70 282.94 Note. a model CAIC < than both independence and saturated CAICs */?<.01. **p<.001. Table 10 Tests of Measurement Invariance for the MBSRQ - Health Evaluation Model A x ' Adf ACFI Invariance? Configural - full model - - - no Female model - - - no Male model - - - yes Young adult model - - - no Middle-aged adult model - - - no Older adult model - - - yes Metric - Male model 34.87** 12 -.03 no Older adult model 21.17* 6 -.02 no Note. Change scores are not calculated for the configural models because they are the baseline models. *p<.01. **p<.001. Body Image 65 As configural invariance was not supported for the full model, the next step in the analysis was to test whether configural invariance was established for any of the five age and gender subgroups (see Table 9). For the female model, configural invariance was not supported, as four of the five fit statistics (i.e., % 2 , R M S E A , NNFI, CAIC) indicated a poor fit of the model. For the male model, configural invariance was supported, with four of the five fit statistics (i.e., R M S E A , NNFI, CFI, CAIC) indicating an acceptable fit of the model. For the young and middle-aged adult models, configural invariance was not supported, with three of the five fit statistics (i.e., % 2 , RMSEA, NNFI) indicating a poor fit of each model. Finally, for the older adult model, configural invariance was supported with three of the five fit statistics (i.e., NNFI, CFI, CAIC) indicating an acceptable fit of the model. Thus, configural invariance was not met for the female, young adult, or middle-aged adult models. For these three groups, these findings suggest that a unidimensional model may not best represent the construct of health evaluation. Thus, observed mean scores should not be compared on this subscale. Configural invariance was supported for the male and older adult models, however, indicating that the same factor configuration did hold for these groups, and that further tests of measurement invariance were warranted. Because configural invariance was supported for the male and older adult models, the next step in the analysis was to test for metric invariance for these models. For the male model, the increase in chi-square from the configural invariance model was significant (A x 2 = 3 4 . 8 7 , p < . 0 0 1 ) and ACFI = - . 0 3 , which indicated that the hypothesis of metric invariance was not tenable. For the older adult model, initial analyses produced a non-positive definite matrix resulting from bad starting values. To circumvent this issue, the model was rerun using the generalized least squares method to obtain better starting values. These new starting values were then used with the maximum likelihood method to produce the final model. The increase Body Image 66 in chi-square from the configural invariance model was significant (A% =21.17,/? = .002) and ACFI = -.02, which indicated that the hypothesis of metric invariance was not tenable. The lack of metric invariance for the male and older adult models indicated that the construct was not perceived in the same manner and that the items were not on equal scale intervals across the groups in these models. Thus, observed mean scores for men and older adults should not be compared on this subscale. To explore why configural invariance did not hold for the female, young adult or middle-aged adult models, exploratory principal components analyses (PCAs) were conducted on this subscale separately for the three female groups to determine whether a non-unidimensional factor pattern emerged for these groups. As configural invariance held for the male model, indicating that a one factor model holds for each male age group, it was not necessary to run a PCA for the young adult or middle-aged adult males. The number of factors to extract for each group was determined by conducting a parallel analysis (Reise, Waller, & Comrey, 2000). This method of extraction was preferred over using the eigenvalue greater than 1.0 rule or the scree test because the former has been found to extract too many factors (Fabrigar, Wegener, MacCallum, & Strahan, 1999) and the latter has been criticized for being too subjective (Kaiser, 1970). In a parallel analysis, a random set of uncorrelated data is generated separately for each group, matching the group data on the number of variables and the sample size. A PCA is then conducted on this randomly generated data and the eigenvalues from the random data are compared to the eigenvalues from the group data. Factors are retained in the group data if their eigenvalue is greater than the largest eigenvalue produced by the random data. A PCA is then rerun for the group data specifying the number of factors to be extracted. The results of the PCA for the three female age groups indicated that one factor Body Image 67 accounted for the variance of the items for all three groups. Although the P C A suggests that a one-factor model holds for these groups, the failure of these groups to show configural invariance may suggest that there may be one or more minor secondary factors accounting for the variance of the items. Health Orientation. Tables 11 and 12 summarize the results of the measurement invariance tests for the Health Orientation subscale of the MBSRQ. The first test of invariance across the six age and gender groups was for configural invariance. Given that three of the five fit statistics (i.e., % 2 , R M S E A , NNFI) did not support the fit of this model, I concluded that configural invariance was not supported for the full model (see Table 11). Thus, it was concluded that the Health Orientation subscale did not show the same factor configuration across the six age and gender groups. As configural invariance was not supported for the full model, the next step in the analysis was to test whether configural invariance was established for any of the five age and gender subgroups (see Table 11). For the female model, configural invariance was not supported, as three of the five fit statistics (i.e., % 2 , R M S E A , NNFI) indicated a poor fit of the model. For the male model, configural invariance was supported, with four of the five fit statistics (i.e., R M S E A , NNFI, CFI, CAIC) indicating an acceptable fit of the model. For the young and middle-aged adult models, configural invariance was not supported, with three of the five fit statistics (i.e., % , R M S E A , NNFI) indicating a poor fit of each model. Finally, for the older adult model, configural invariance was supported with four of the five fit statistics (i.e., R M S E A , NNFI, CFI, CAIC) indicating an acceptable fit of the model. Thus, configural invariance was not met for the female, young adult, or middle-aged adult models. These findings suggest that a unidimensional model may not best represent the construct of Health Orientation for these groups. Configural invariance was shown for the male and older adult Table 11 Goodness-of-Fit Indices for the MBSRQ - Health Orientation Model x 2 df R M S E A NNFI CFI Model CAIC Independence CAIC Saturated CAIC Configural - full model 435.69** 120 .12 .88 .92 1247.23a 4318.04 1756.96 Female model 289.61** 60 .12 .87 .92 687.89 a 2768.43 834.56 Male model 146.07** 60 .10 .90 .93 487.20 a 1513.54 760.09 Young adult model 193.22** 40 .13 .87 .91 450.21 a 1869.61 525.92 Middle-aged adult model 141.48** 40 .12 .87 .91 371.063 1260.27 503.02 Older Adult model 100.98** 40 .10 .91 .94 317.873 1134.11 484.80 Metric - Male model 177.05** 76 .10 .92 .93 408.75 a 1623.26 760.09 Older adult model 123.02** 48 .10 .89 .91 286.33 a 953.61 484.80 Scalar - Male model 255.66** 88 .12 .87 .87 578.58a 1513.54 760.09 Note. a model CAIC < than both independence and saturated CAICs *p<M. **p<.001. Table 12 Tests of Measurement Invariance for the MBSRQ - Health Orientation Model A x ' Adf ACFI Invariance? Configural - full model - - - no Female model - - - no Male model - - - yes Young adult model - - - no Middle-aged adult model - - - no Older adult model - - - yes Metric - Male model 30.99 16 0 yes Older adult model 22.04* 8 -.03 no Scalar - Male model 109.59** 12 -.06 no Note. Change scores are not calculated for the configural models because they are the baseline models. *p<.Q\. **/?<.001. Body Image 70 models, however, which indicated that the same factor configuration did hold for these groups, and that further tests of measurement invariance were warranted. For the male and older adult models, the next step in the analysis was to test for metric invariance for these groups. For the male model, the increase in chi-square from the configural invariance model was not significant (A)[ 2 = 30.99,/? = .01) and ACFI = 0, which indicated that the hypothesis of metric invariance was tenable. Thus, tests of scalar invariance can be conducted for this group. For the older adult model, the increase in chi-square from the configural invariance model was significant (A % 2 = 22.04,p = .005) and ACFI = -.03, which indicated that the hypothesis of metric invariance was not tenable. This finding indicates that older adult men and women did not perceive the construct of health orientation in the same manner and that the items were not on equal scale intervals across these groups. Thus, observed mean scores should not be compared on this subscale for women and older adults. As metric invariance was supported for the male model, the final test of measurement invariance examined for this group was for scalar invariance. The increase in chi-square from the configural invariance model was significant ( A x = 109.59,/? < .001) and ACFI = -.06, which indicated that the hypothesis of scalar invariance was not tenable. Thus, measurement bias cannot be ruled out when comparing across groups. Therefore, observed mean scores for men should not be compared across age groups on this subscale. To explore why configural invariance did not hold for the female, young adult, or middle-aged adult models, exploratory PCAs were conducted on this subscale separately for each of the three female age groups to determine whether a non-unidimensional factor pattern emerged for these groups. Because configural invariance held for the male model, indicating that a one factor model holds for the three male groups, it was not necessary to run a P C A for the young adult or middle-aged adult males. The results of the P C A indicated that two factors Body Image 71 accounted for the variance of the items for the young adult and middle-aged adult women, whereas one factor accounted for the variance of the items for the older adult women. For both the young adult and middle-aged adult women, items 8, 29, and 52 loaded on one factor, and items 18, 19, 28, and 38 loaded on a second factor. Item 9 loaded on both factors for the young adult women and on the latter factor for the middle-aged adult women. Illness Orientation. Tables 13 and 14 summarize the results of the measurement invariance tests for the Illness Orientation subscale of the MBSRQ. The first test of invariance across the six age and gender groups was for configural invariance. Although the significant chi-square and R M S E A indicated a poor fit of the model, the remaining three fit statistics (i.e., NNFI, CFI, CAIC) indicated an acceptable fit of the model and thus I concluded that, although this model did not provide a great fit to the data, it did provide an acceptable fit and further tests of measurement invariance were warranted (see Table 13). Thus, it was concluded that the Illness Orientation subscale showed the same factor configuration across the six age and gender groups. As configural invariance was supported for the full model, the next step in the analysis was to test for metric invariance for the full model. For this model, initial analyses produced a non-positive definite matrix resulting from bad starting values. The model was rerun using the generalized least squares method to obtain better starting values that were then used with the maximum likelihood method to produce the final model. The increase in chi-square from the configural invariance model was significant ( A x - 207.43, p < .001) and ACFI = -.03, which indicated that the hypothesis of metric invariance was not tenable. Thus, metric invariance was not shown for the full model. This finding indicates that the strength of the relationship between each item and factor was significantly different between the groups. Table 13 Goodness-of-Fit Indices for the MBSRQ - Illness Orientation Model x 2 df R M S E A NNFI CFI Model CAIC Independence CAIC Saturated CAIC Configural - full model 110.58** 30 .12 .90 .95 604.08a 1980.45 732.07 Metric - full model 189.87** 54 .11 .91 .92 482.27a 1989.44 732.14 Female model 101.91** 24 .14 .92 .93 270.32a 1330.83 347.79 Male model 79.69** 24 .13 .86 .89 225.45 a 636.07 316.70 Young adult model 72.88** 14 .13 .89 .92 196.41a 837.82 219.13 Middle-aged adult model 45.20** 14 .14 .91 .93 155.60a 567.78 209.59 Older adult model 22.51 14 .06 .97 .98 129.52 s 550.08 202.10 Scalar - Female model 143.80** 31 .12 .91 .90 371.04 1330.83 347.79 Young adult model 74.86** 18 .11 .92 .92 241.90 837.82 219.13 Older adult model 105.32** 18 .15 .80 .82 246.82 567.78 209.59 Note. a model CAIC < than both independence and saturated CAICs */?<.01. **/?<.001. Table 14 Tests of Measurement Invariance for the MBSRQ - Illness Orientation Model A x ' Adf ACFI Invariance? Configural - full model - - - yes Metric - full model 207.43** 24 -.03 no Female model 15.24 9 -.01 yes Male model 55.82** 9 -.09 no Young adult model 7.43 4 -.01 yes Middle-aged adult model 21.28 4 -.04 no Older Adult model 1.34 4 0 yes Scalar - Female model 57.13** 16 -.04 no Young adult model 9.41 8 -.01 yes Older adult model 84.15** 8 -.16 no Note. Change scores are not calculated for the configural model because this model is the baseline model. *p<.Q\. **/?<.001. Body Image 74 Because metric invariance was not supported for the full model, the next step was to assess whether metric invariance was established for any of the five age and gender subgroups. The results are summarized in Table 14. Metric invariance was not shown for the male model or middle-aged adult model, indicating that the construct of illness orientation was not perceived in the same manner and that the items were not on equal scale intervals across these groups. Thus, observed mean scores for men and middle-aged adults should not be compared on this subscale. Metric invariance was met for the female, young adult, and older adult models, however, indicating that tests of scalar invariance could be conducted for these groups. The final measurement invariance test conducted was for scalar invariance for the female, young adult, and older adult models. As indicated in Table 14, scalar invariance was not met for the young adult model or older adult model, suggesting that measurement bias exists that caused these groups to respond differently to the items and that observed mean scores for these groups should not be compared on this subscale. Scalar invariance was shown for the female model, however, suggesting that women responded to the items in a similar way and that comparisons of observed scores may be examined and interpreted across age groups. Body Areas Satisfaction Subscale. Tables 15 and 16 summarize the results of the measurement invariance tests for the Body Areas Satisfaction Subscale of the MBSRQ. The first test of invariance across the six age and gender groups was for configural invariance. Given that three of the five fit statistics (i.e., % , RMSEA, NNFI) did not support the fit of this model, I concluded that configural invariance was not supported for the full model (see Table 15). Thus, it was concluded that the Body Areas Satisfaction subscale did not show the same factor configuration across the six age and gender groups. Table 15 Goodness-of-Fit Indices for the MBSRQ - Body Areas Satisfaction Model x 2 df R M S E A NNFI CFI Model CAIC Independence CAIC Saturated CAIC Configural - full model 745.59** 162 .14 .87 .90 1684.81 a 6601.92 2196.42 Female model 448.58** 81 .14 .88 .91 925.16 s 4242.87 1043.36 Male model 297.01** 81 .14 .86 .89 678.46a 2318.46 950.11 Young adult model 262.28** 54 .13 .90 .92 565.59a 2888.28 657.40 Middle-aged adult model 263.33** 54 .15 .85 .88 533.82a 2005.91 628.78 Older adult model 219.98** 54 .14 .85 .89 463.82a 1646.91 606.29 Metric - Young adult 280.24** 62 .10 .92 .93 522.93a 2792.80 657.40 Scalar - Young adult 355.88** 70 .13 .89 .89 687.13 2888.28 657.40 Note. a model CAIC < than both independence and saturated CAICs *p<M. **/?< .001. Table 16 Tests of Measurement Invariance for the MBSRQ - Body Areas Satisfaction Model A x ' Adf ACFI Invariance? Configural - full model - - - no Female model - - - no Male model - - - no Young adult model - - - yes Middle-aged adult model - - - no Older adult model - - - no Metric - Young adult 17.96 8 0 yes Scalar - Young adult 93.60** 8 -.09 no Note. Change scores are not calculated for the configural models because these models are the baseline models. *p<.0\. **/?<.001. Body Image 77 As configural invariance was not supported for the full model, the next step in the analysis was to test whether configural invariance was established for any of the five age and gender subgroups (see Table 15). For the female model, configural invariance was not supported, as three of the five fit statistics (i.e., % 2 , RMSEA, NNFI) indicated a poor fit of the model. For the male, middle-aged, and older adult models, configural invariance was not supported, with four of the five fit statistics (i.e., x 2 , R M S E A , NNFI, CFI) indicating a poor fit of each model. Finally, for the young adult model, configural invariance was supported, with three of the five fit statistics (i.e., NNFI, CFI, CAIC) indicating an acceptable fit of the model. Thus, configural invariance was not met for the female, male, middle-aged adult, or older adult models. These findings suggest that a unidimensional model may not best represent the construct of body areas satisfaction for these groups. Configural invariance was supported for the young adult model, however, indicating that further tests of measurement invariance were warranted for this group. Because configural invariance was supported for the young adult model, the next step in the analysis was to test for metric invariance for this model. Initial analyses produced a non-positive definite matrix resulting from bad starting values. The model was rerun using the generalized least squares method to obtain better starting values that were then used with the maximum likelihood method to produce the final model. The increase in chi-square from the configural invariance model was not significant (A x 2 = 17 .96,/? = .02) and ACFI = 0 , which indicated that the hypothesis of metric invariance was tenable. Thus, metric invariance was established for the young adult model and further tests of scalar invariance could be conducted. Because metric invariance was supported for the young adult model, the final test of measurement invariance examined for this model was for scalar invariance. The increase in Body Image 78 chi-square from the configural invariance model was significant (A % 2 = 93.60,/? < .001) and ACFI = -.09, which indicated that the hypothesis of scalar invariance was not tenable. Thus, scalar invariance was not established for the young adult model, suggesting that measurement bias exists that caused the young adult men and women to respond differently to the items and that observed mean scores for young adults should not be conducted on this subscale. To explore why configural invariance did not hold for the female, male, middle-aged adult, or older adult models, exploratory PCAs were conducted on this subscale separately for each group to determine whether a non-unidimensional factor pattern emerged for these groups. The results of the P C A indicated that two factors accounted for the variance of the items for the middle-aged adult group of women and that one factor accounted for the variance of the items for the young and older adult groups of women and for all three groups of men. For the middle-aged adult group of women, items 61, 62, and 68 loaded on one factor, and items 63, 64, 65, 66, 67, and 69 loaded on a second factor. Although the PCA suggests that a one-factor model holds for the three male groups, the failure of the male model to shown configural invariance suggests that there may be one or more minor secondary factors accounting for the variance of the items. Overweight Preoccupation. Tables 17and 18 summarize the results of the measurement invariance tests for the Overweight Preoccupation subscale of the MBSRQ. The first test of invariance across the six age and gender groups was for configural invariance. Although the significant chi-square and R M S E A indicated a poor fit of the model, the remaining three fit statistics (i.e., NNFI, CFI, CAIC) indicated an acceptable fit of the model and thus I concluded that, although this model did not provide a great fit to the data, it did provide an acceptable fit and further tests of measurement invariance were warranted (see Table 17). Thus, it was concluded that the Overweight Preoccupation subscale showed the same factor configuration Table 17 Goodness-of-Fit Indices for the MBSRQ - Overweight Preoccupation Model x 2 df R M S E A NNFI CFI Model CAIC Independence CAIC Saturated CAIC Configural - full model 41.77** 12 .11 .92 .97 432.47a 1312.27 488.09 Metric - full model 76.20** 31 .08 .95 .96 311.41 a 1312.27 488.09 Scalar - full model 155.72** 42 .11 • .91 .89 491.68 1312.27 488.09 Female model 101.80** 18 .13 .89 .89 285.37 874.25 231.86 Male model 42.89** 18 .11 .92 .92 209.803 419.97 211.14 Young adult model 18.25 10 .06 .98 .98 149.96 548.88 146.09 Middle-aged adult model 12.82 10 .04 .99 .99 138.378 414.66 139.73 Older Adult model 35.64** 10 .13 .88 .90 158.27 321.71 134.73 Note. a model CAIC < than both independence and saturated CAICs *p<S\. **p<.001. Table 18 Tests of Measurement Invariance for the MBSRQ - Overweight Preoccupation Model A x ' Adf ACFI Invariance? Configural - full model - - - yes Metric - full model 34.43 19 -.01 yes Scalar - full model 113.95** 30 -.08 no Female model 73.53** 12 -.08 no Male model 29.38* 12 -.06 no Young adult model 10.71 6 -.01 yes Middle-aged adult model 12.27 6 0 yes Older Adult model 7.39 6 -.01 yes Note. Change scores are not calculated for the configural model because this model is the baseline model. * < .01. **/><.001. Body Image 81 across the six age and gender groups. As configural invariance was supported for the full model, the next step in the analysis was to test for metric invariance for the full model. For this model, initial analyses produced a non-positive definite matrix resulting from bad starting values. The model was run using the generalized least squares method to obtain better starting values that were then used with the maximum likelihood method to produce the final model. The increase in chi-square from the configural invariance model was not significant ( A x = 34.43, p = .02) and ACFI = -.01, which indicated that the hypothesis of metric invariance was tenable. Thus, the strength of the relationship between each item and factor was not statistically different between the groups and further tests of measurement invariance were warranted. As metric invariance was supported for the full model, the next step in the analysis was to test for scalar invariance for the full model. The increase in chi-square from the configural invariance model was significant ( A x = 113.95,/? < .001) and ACFI = -.08, which indicated that the hypothesis of scalar invariance was not tenable. This finding suggests that measurement bias exists that caused the groups to respond differently to the items. As scalar invariance was not supported for the full model, the final step was to assess whether scalar invariance was established for any of the five age and gender subgroups. The summary of results is presented in Table 18. Scalar invariance was not met for the female or male models. These findings suggest that measurement bias cannot be ruled out when comparing women or men across age groups and thus observed mean scores for these groups should not be compared on this subscale. Scalar invariance was shown for the young, middle, and older adult models, however, indicating that these groups responded to the items in the same way and that differences in observed scores on this subscale may be conducted and interpreted across these groups. Body Image 82 Gender and Age Differences in Mean Scores on the MBSRQ Subscales Table 19 presents the means and standard deviations for each subscale of the MBSRQ. For those subgroups that met the hypothesis of scalar invariance, univariate one-way analyses of variance (ANOVAs) were conducted with gender or age group as the independent variable and the subscale mean score as the dependent variable. Two-way A N O V A s with gender and age group could not be conducted because scalar invariance was not achieved for the full model for any subscale, and was only achieved for select subgroups. Table 19 Means (Standard Deviations) for the Subscales of the MBSRQ Young Middle-aged Older M(SD) M(SD) M(SD) Appearance Evaluation Men 3.69 (.70) 3.48 (.70) 3.49 (.50) Women 3.43 (.82)a 3.33 (.89)a 3.23 (.77)a Appearance Orientation Men 3.23 (.62) 3.20 (.66)a 3.49 (.50) Women 3.48 (.60) 3.46 (.63)a 3.49 (.63) Fitness Orientation Men 3.76(.74) a b 3.45 (.66) a c 3.43 (.67) a d Women 3.47 (.71)b 3.48(.72) c 3.40 (.70)d Health Evaluation Men 3.90.(.73) 3.77 (.60) 3.84 (.62) Women 3.72 (.72) 3.86 (.76) 3.81 (.82) Health Orientation Men 3.46 (.70) 3.56 (.66) 3.72 (.60) Women - - 3.95 (.60) Body Image 83 Illness Orientation Men 2.99 (.75)a 3.18 (.78) 3.34 (.80) Women 3.26 (.77)a 3.28 (.76) 3.27 (.79) Body Areas Satisfaction Men 3.60 (.68) 3.50 (.61) 3.62 (.49) Women 3.39 (.64) - 3.36 (.65) Overweight Preoccupation Men 1.89 (.74)a 2.34 (.82)" 2.13 (.72)c Women 2.60 (.89)a 2.69 (.91)b 2.50 (.89)c Note. Means with the same subscript within each subscale exhibit scalar invariance and can be compared; missing means and standard deviations represent subscales that did not exhibit unidimensionality and thus the mean score should not be used. Appearance Evaluation. As scalar invariance was established for the female model, a univariate A N O V A was conducted to examine any age-related differences. Results indicated a significant small effect for age group, F (2, 837) = 4.27,/? = .014, eta-sq. = .014 . Follow-up post poc analyses indicated that young adult females reported significantly greater satisfaction with appearance than older adult females. There were no significant differences between the middle-aged women and either the young women or older women. Appearance Orientation. As scalar invariance was established for the middle-aged adults, a univariate A N O V A was conducted to examine any gender differences. Results indicated a significant small effect for gender with females reporting greater investment in appearance than males, F (1, 396) = 14.77,/? < .001 , eta-sq. = .036. 2 Effect sizes are reported in addition to the statistical test results to indicate whether an effect is non-trivial or not (Zumbo & Hubley, 1998). Kirk's (1996) criteria for interpreting effect size are as follows: small effect = .010 to .058, medium effect = .059 to .137, and large effect = >.137. Kirk's criteria are for omega-sq.; however, these criteria may be appropriately applied to interpreting partial eta-squared which is a similar measure of strength of association. 3 All p values will be reported as exact values, except in that case where p is less that .00,1 in which case p will be reported as < .001. Body Image 84 Fitness Orientation. As scalar invariance was established for four of the five subgroups, a series of univariate A N O V A s were conducted to examine age differences for men, and gender differences for young, middle-aged, and older adults. Results indicated a small significant effect for age group for the men, F (2, 419) = 10.94,/? < .001, eta-sq. = .050. Follow-up post hoc analyses indicated that young men reported significantly higher investment in fitness than both middle-aged men and older adult men. There were no significant differences between the middle-aged and older men. Results for the gender differences indicated a significant small effect for gender for young adults, with young men reporting significantly more investment in fitness than young women, F (1, 547) = 19.66 /? < .001, eta-sq. = .035. There were no significant main effects for gender for middle-aged adults, F (1, 396) = .19, n.s., eta-sq. < .001, or older adults, F(\, 313) = .134, n.s., eta-sq. < .001. Illness Orientation. As scalar invariance was established only for young adults, a univariate A N O V A was conducted to examine any gender differences. Results indicated a significant small effect for gender, with young women reporting a greater awareness of physical symptoms and an increased likelihood of seeking medical attention than young men, F(l, 549)= 15.61,/?<.001, eta-sq. =.028. Overweight Preoccupation. As scalar invariance was established for the young, middle-aged, and older adult groups, a series of univariate A N O V A s were conducted to examine any gender differences in each of the three age groups. Results indicated significant effects for gender for each of the age groups, with young, F (1, 549) = 87.57,/? < .001, eta-sq. = .138, middle-aged, F(\, 396) = 14.47,/? < .001, eta-sq. = .035, and older F ( l , 313) = 13.39, p < .001, eta-sq. = .04, women reporting significantly higher scores than their male counterparts. There was a large effect size for the group of young adults and small effect sizes for the middle-aged and older adults. Body Image 85 Scalar invariance was not established for any subgroups for the Health Evaluation, Health Orientation, and Body Areas Satisfaction subscales, so no group comparisons are made using these subscales. Invariance of the Appearance Schemas Inventory-Revised The ASI-R is composed of two subscales - Self-Evaluative Salience and Motivational Salience. Thus, invariance testing was conducted based on the assumption that a similar two-factor structure would exist across all groups. However, Cash et al. (2004) also assert that the ASI-R can be summed to a composite score. Thus, a one-factor model was also tested for this scale. The results are presented first for the two-factor model and then for the one-factor model. Two-factor model. Tables 20 and 21 summarize the results of the two-factor measurement invariance tests for the ASI-R. The first test of invariance across the six age and gender groups was for configural invariance. Although the chi-square for this analysis was significant, the four other fit statistics (i.e., RMSEA, NNFI, CFI, CAIC) indicated an acceptable fit of the model (see Table 20). Thus, it was concluded that the ASI-R showed the same factor configuration across the six age and gender groups, and that further tests of measurement invariance were warranted. As configural invariance was supported for the full model, the next step in the analysis was to test for metric invariance for the full model. The increase in chi-square from the configural invariance model was significant (A % 2 = 748.52,/? < .001) and ACFI = -.02, which indicated that the hypothesis of metric invariance was not tenable. This finding indicates that the strength of the relationship between each item and factor was significantly different between the groups. Table 20 Goodness-of-Fit Indices for the ASI-R Two-Factor Model Model x 2 df R M S E A NNFI CFI Model CAIC Independence CAIC Saturated CAIC Configural - full model 3001.59** 1018 .10 .90 .91 4979.77a 25387.04 10244.93 Metric - full model 3750.11** 1118 .12 .88 .89 5376.29a 25387.04 10244.93 Female model 2465.27** 548 .12 .87 .87 3518.43a 15993.57 4564.50 Male model 1246.29** 548 .10 .91 .92 1901.363 9303.47 4435.36 Young adult model 1095.59** 358 .09 .92 .92 1628.12 s 10380.64 3068.64 Middle-aged adult model 895.33** 358 .09 .93 .94 1355.193 9065.36 2927.93 Older adult model 1196.24** 358 .13 .83 .84 1750.41a 5805.89 2829.36 Reduced female model 1091.81** 358 .08 .93 .94 1591.223 12206.69 3124.53 Scalar - Male model 1366.86** 581 .11 .91 .91 2246.343 9303.47 4435.36 Young adult model 1196.07** 375 .10 .91 .92 1946.803 10380.64 3068.64 Middle-aged adult model 993.60** 375 .10 .93 .93 1647.64 3 9065.37 2927.93 Reduced female model 1181.82** 375 .09 .93 .93 1865.953 12206.69 3124.53 Note. 3 model CAIC < than both independence and saturated CAICs *p<.0\. **p<.001. Table 21 Tests of Measurement Invariance for the ASI-R Two-Factor Model Model A x ' Adf ACFI Invariance? Configural - full model - - - yes Metric - full model 748.52** 100 -.02 no Female model 857.92** 40 -.05 no Male model 49.04 40 0 yes Young adult model 18.21 20 0 yes Middle-aged adult model 49.09** 20 0 yes Older adult model 315.42** 20 -.05 no Reduced female model 15.00 20 0 yes Scalar - Male model 169.61** 73 -.01 yes Young adult model 118.69** 37 0 yes Middle-aged adult model 147.36** 17 -.01 yes Reduced female model 105.04** 37 -.01 yes Note. Change scores are not calculated for the configural model because this model is the baseline model. *p<M. **p<.001. Body Image 88 As metric invariance was not supported for the full model, the next step was to assess whether metric invariance was established for any of the five age and gender subgroups. These results are presented in Table 21. Metric invariance was not shown for the female or older adult models, indicating that these groups did not perceive the construct of appearance investment in the same manner and that the items were not on equal scale intervals across the groups. Thus, observed mean scores for these groups should not be compared on this scale. Metric invariance was supported for the male model, young adult model, and middle-aged adult model, however, indicating that further tests of measurement invariance can be conducted for these groups. As the male, young adult, and middle-aged adult models met the hypothesis of metric invariance, the final measurement invariance test conducted for these three groups was for scalar invariance. As indicated in Table 21, scalar invariance was shown for all three models. These findings suggest that these groups responded to the items in the same way and that comparisons of observed scores can be examined and interpreted across these groups. Given that scalar invariance was not met for the female model (i.e., young, middle-aged, and older adult women), but scalar invariance was met for the young and middle-aged adult models (i.e., men and women within those age groups), post hoc invariance analyses were conducted to test whether the young and middle-aged groups of women were invariant on this scale. This model was referred to as the reduced female model. These results are summarized in Tables 20 and 21. The results of the metric invariance test indicated that the increase in chi-square from the configural invariance model was not significant (A % 2 = 15.00, p = .78) and ACFI = 0, which indicated that the hypothesis of metric invariance was tenable. The results of the scalar invariance test indicated that the increase in chi-square from the configural invariance model was significant (A % 2 = 105.04,/? < .001), which indicated that the Body Image 89 hypothesis of scalar invariance was not tenable. However, ACFI = -.01, which indicated that the hypothesis of scalar invariance was tenable. Following my arguments from other analyses, I concluded that scalar invariance was met for the reduced female model. This finding indicates that young and middle-aged women responded to the items in the same way and the comparisons of observed scores can be examined and interpreted across these groups. One-factor model. Table 22 summarizes the results of the one-factor measurement invariance tests for the ASI-R. The first test of invariance across the six age and gender groups was for configural invariance. As four of the five fit statistics (i.e., x , R M S E A , NNFI, CFI) indicated a poor fit of the model, I concluded that configural invariance was not supported for the full model. This finding indicates that a single factor configuration did not hold for all six age and gender groups. As configural invariance was not supported for the full model, the next step in the analysis consisted of testing for configural invariance for each of the five age and gender subgroups. For the female, male, young adult, and older adult models, configural invariance was not supported, with four of the five fit statistics (i.e., x , R M S E A , NNFI, CFI) indicating a poor fit of each model. For the middle-aged adult model, configural invariance was not supported, with three of the five fit statistics (i.e., % 2 , R M S E A , NNFI) indicating a poor fit of the model. Thus, configural invariance was not supported for any of the five subgroups. These findings indicate that a one-factor model did not best represent the construct of dysfunctional schematic appearance investment. This suggests that a total score should not be used for this scale. Table 22 Goodness-of-Fit Indices for the ASI-R One-Factor Model Model x 2 df R M S E A NNFI CFI Model CAIC Independence CAIC Saturated CAIC Invariance? Configural - full model 3842.61** 1020 .14 .86 .88 6969.85a 25387.04 10244.93 no Female model 2383.55** 510 .14 .86 .87 4199.68a 15993.57 4864.50 no Male model 1459.05** 510 .13 .87 .89 2590.16a 9303.47 4435.36 no Young adult model 1432.09** 340 .13 .87 .89 2416.103 10380.64 3068.64 no Middle-aged adult model 1204.00** 340 .14 .89 .90 2174.37a 9065.36 2927.93 no Older Adult model 1206.52** 340 .15 .81 .83 2109.11a 5805.89 2829.36 no Note. a model CAIC < than both independence and saturated CAICs *p<.§\. **/?<.001. Body Image 91 Gender and age differences in mean scores on the ASI-R Subscales. Table 23 presents the means and standard deviations on the ASI-R. As scalar invariance was met for all subgroups except for the older women, a series of univariate one-way A N O V A s were conducted to examine gender differences between young and middle-aged adults, age differences among all three groups of men, and age differences between young and middle-aged women. Separate analyses were conducted for the Self-Evaluative Salience (SES) subscale and the Motivational Salience (MS) subscale. Table 23 Means (Standard Deviations) for the Subscales of the ASI-R Young Middle-aged Older M(SD) M(SD) M(SD) Self-Evaluative Salience Men 2.95 (.72)ac 2.82 (.71) a d 2.59 (.60)a Women 3.25 (.69) b c 3.01 (.73) b d 2.71 (.65) Motivational Salience Men 3.39 (.67) a c 3.27 (.71) a d 3.21 (.67)a Women 3.62 (.63) b c 3.52 (.68) b d 3.50 (.68) Note. Means with the same subscript exhibit scalar invariance with each subscale and can be compared. Self-Evaluative Salience. For the SES subscale, results indicated significant small effects for gender for the young, F{\, 547) = 22.98, p < .001, eta-sq. = .040, and middle-aged adults F (1, 396) = 5.92, p = .015, eta-sq. — .015, with women reporting a greater investment in appearance than men. A significant small effect for age group was found for men, F (2,419) = 9.23, p < .001, eta-sq. = .042. Follow-up post hoc analyses indicated that the group of older Body Image 92 men reported significantly less investment in appearance than both the young and middle-aged men. There were no significant differences between the young and middle-aged men. Finally, a small significant effect for age group was also found between the young adult and middle-aged women, with the group of young women reporting significantly greater investment in appearance than the middle-aged women, F (1, 629) = 17.92, p < .001, eta-sq. = .028. Motivational Salience. For the MS subscale, results indicated significant small effects for gender for the young, F(l, 547) = 15.09,/? < .001, eta-sq. = .027, and middle-aged adults, F(\, 396) = 11.12,/? = .001, eta-sq. = .027, with women reporting significantly greater engagement in appearance-management behaviors than their male counterparts. In looking at age group effects for men, no significant differences were found, F (2, 419) = 2.67,/? = .070, eta-sq. = .013. Finally, no significant age differences were found between the young and middle-aged women, F(l, 629) = 3.75,/? = .053, eta-sq. = .006. Invariance of the Body Image Quality of Life Inventory Two separate analyses were conducted on this scale with (1) all items remaining in the scale, but those participants with missing data for items 6, 11, and 12 removed, and (2) items 6, 11, and 12 removed from the scale. The results will be reported first for the full scale with all 19 items and then for the reduced scale with items 6, 11, and 12 removed. Both scales examined whether a single factor structure exists across all groups. Full Scale. Tables 24 and 25 summarize the results of the measurement invariance tests for the BIQLI. The first test of invariance across the six age and gender groups was for configural invariance. Although the significant chi-square and R M S E A indicated a poor fit of the model, the remaining three fit statistics (i.e., NNFI, CFI, CAIC) indicated an acceptable fit of the model and thus I concluded that, although this model did not provide a great fit to the data, it did provide an acceptable fit and further tests of measurement invariance were Table 24 Goodness-of-Fit Indices for the BIQLI-Full Scale x2 df R M S E A NNFI CFI Model Independence Saturated Model CAIC CAIC CAIC Configural - full model 5266.38** 912 .15 .92 .93 7059.35a 60698.86 9230.26 Metric - full model 5591.93** 1007 .15 .92 .92 6616.013 60698.86 9230.26 Scalar - full model 5638.72** 1092 .14 .93 .92 6939.98a 60741.70 9230.26 Note. a model CAIC < than both independence and saturated CAICs *p< .01. **p< .001. Table 25 Tests of Measurement Invariance for the BIQLI- Full Scale Model A % l Adf ACFI Invariance? Configural - full model - - - yes Metric - full model 325.55** 95 -.01 yes Scalar-full model 372.34** 180 -.01 yes Note. Change scores are not calculated for the configural model because this model is the baseline model. *p<.01. **/?<.001. Body Image 95 warranted (see Table 24). Thus, it was concluded that the Body Image Quality of Life Inventory showed the same factor configuration across the six age and gender groups. As configural invariance was supported for the full model, the next step in the analysis was to test for metric invariance for the full model. The increase in chi-square from the configural invariance model was significant (A % = 325.55, p < .001), which indicated that the hypothesis of metric invariance was not tenable. However, ACFI = -.01, which indicated that the hypothesis of metric invariance was tenable. Following my arguments from other analyses, I concluded that metric invariance was shown for the full model. Thus, the strength of the relationship between each item and factor was not statistically different between the groups. As metric invariance was supported for the full model, the next step in the analysis was to test for scalar invariance for the full model. The increase in chi-square from the configural invariance model was significant (A % 2 = 372.34,/? < .001), which indicated that the hypothesis of scalar invariance was not tenable. However, ACFI = -.01, which indicated that the hypothesis of scalar invariance was tenable. Following my arguments from other analyses, I concluded that scalar invariance was shown for the full model. This finding suggests that all six age and gender groups responded to the items in the same way and that comparisons of observed scores can be examined and interpreted across all groups. Reduced scale. Tables 26 and 27 summarize the results for the measurement invariance tests for the reduced BIQLI. The conclusions for the reduced scale mirrored those for the full scale. The first test of invariance across the six age and gender groups was for configural invariance. Although the significant chi-square and R M S E A indicated a poor fit of the model, the remaining three fit statistics (i.e., NNFI, CFI, CAIC) indicated an acceptable fit of the model and thus I concluded that, although this model did not provide a great fit to the data, it did provide an acceptable fit and further tests of measurement invariance were Table 26 Goodness-of-Fit Indices for the BIQLI- Reduced Scale I1 df RMSEA NNFI CFI Model Independence Saturated Model CAIC CAIC CAIC Configural - full model 3783.43** 624 .15 .91 .93 5258.57a 43953.70 6636.12 Metric - full model 4076.59** 704 .15 .92 .92 4894.40 a 43953.70 6636.12 Scalar - full model 4121.62** 774 .14 .93 .92 5210.16s 43999.94 6636.12 Note. a model CAIC < than both independence and saturated CAICs */?< .01. **p<.001. Table 27 Tests of Measurement Invariance for the BIQLI- Reduced Scale Model A p Adf ACFI Invariance? Configural - full model . . . v e s Metric - full model 293.16** 80 -.01 yes Scalar-full model 338.19** 150 -.01 yes Note. Change scores are not calculated for the configural model because this model is the baseline model. *p< .01. **/?<.001. Body Image 98 warranted (see Table 26). Thus, it was concluded that the Body Image Quality of Life Inventory showed the same factor configuration across the six age and gender groups. As configural invariance was supported for the full model, the next step in the analysis was to test for metric invariance for the full model. The increase in chi-square from the configural invariance model was significant (A x 2 = 293.16,/? < .001), which indicated that the hypothesis of metric invariance was not tenable. However, ACFI = -.01, which indicated that the hypothesis of metric invariance was tenable. Following my arguments from other analyses, I concluded that metric invariance was shown for the full model. Thus, the strength of the relationship between each item and factor was not statistically different between the groups. As metric invariance was supported for the full model, the next step in the analysis was to test for scalar invariance for the full model. The increase in chi-square from the configural invariance model was significant (A x 2 = 338.19,/? < .001), which indicated that the hypothesis of scalar invariance was not tenable. However, ACFI = -.01, which indicated that the hypothesis of scalar invariance was tenable. Following my arguments from other analyses, I concluded that scalar invariance was shown for the full model. This finding suggests that all six age and gender groups responded to the items in the same way and that comparisons of observed scores can be examined and interpreted across all groups. Gender and age differences in mean scores on the BIQLI Full Scale. Table 28 presents the means and standard deviations for the full BIQLI. As the full model met the hypothesis of scalar invariance, a 3 (age group) x 2 (gender) A N O V A was conducted with the mean score obtained on the BIQLI as the dependent variable4. There 4 When conducting the comparison tests for this scale, those participants who had missing data for items 6, 11, and/or 12 were included in the data set. The mean score for these individuals was calculated by taking the mean of 16 items instead of all 19 items. Analyses run without the participants who had missing data for these items resulted in identical conclusions. Body Image 99 was no significant interaction, F (2, 1256) = 1.36,/? = .297, eta-sq. = .002. A statistically significant, but not practically significant, main effect was found for gender, with men ( M = 1.11 SD = .96) reporting higher body image quality of life than women (M = .91, SD = 1.12), F(1, 1256) = 7.45,/? = .006, eta-sq. = .006. A small significant main effect was also found for age group, F (2, 1256) = 12.05,/? < .001, eta-sq. = .019. Follow-up post hoc analyses indicated that older adults ( M = 1.24, SD = .98) reported a significantly higher body image quality of life than young adults (M = .96, SD = .99) who reported a significantly higher body image quality of life than middle-aged adults (M = .79, SD = 1.21). Table 28 Means (Standard Deviations) for the Full BIQLI Young Middle-aged Older M(SD) M(SD) M(SD) Men 1.12 (.88) .97(1.09) 1.26 (.88) Women .89(1.03) .70(1.25) 1.23 (1.02) Reduced Scale. Table 29 presents the means and standard deviations for the reduced BIQLI. As the full model met the hypothesis of scalar invariance, a 3 (age group) x 2 (gender) A N O V A was conducted with the mean score obtained on the reduced BIQLI as the dependent variable. There was no significant interaction, F (2, 1256) = 1.53,/? = .216, eta-sq. = .002. A statistically significant, but not practically significant, main effect was found for gender, with men (M = 1.10 SD - .95) reporting higher body image quality of life than women (M = .91, SD = 1.11), F (1, 1256) = 6.85,/? = .009, eta-sq. = .005. A small significant main effect was also found for age group, F (2, 1256) = 14.03,/? < .001, eta-sq. = .022. Follow-up post hoc analyses indicated that older adults (M = 1.26, SD = .98) reported a significantly higher body image quality of life than young adults (M = .96, SD = .98) who reported a significantly higher Body Image body image quality of life than middle-aged adults (M = .78, SD = 1.18). Table 29 Means (Standard Deviations) for the Reduced BIQLI Young Middle-aged Older M(SD) M(SD) M(SD) Men 1.12 (.88) .96(1.08) 1.26 (.89) Women .88(1.02) .69(1.23) 1.27(1.02) Body Image 101 Chapter Five Discussion It has been common practice for researchers wishing to study group differences on body image to select the body image measure in which they are most interested and then administer this measure to the groups they want to compare. The results obtained would then be used to examine differences in observed scores. Any significant differences that emerged between groups would be explained as resulting from actual differences between the groups. However, this approach is problematic because the majority of body image measures have been developed with, and for, young adult populations. Using these measures outside of this population, without evidence of measurement invariance, may prevent a clear-cut interpretation of the results. That is, an alternative explanation is introduced in that any differences found between groups may result from differences in the way the items are interpreted by members of each group. This may possibly result in different latent variables being measured in each of the groups, making comparisons between groups inappropriate. For researchers who are interested in exploring group differences, an important first step is to ensure that the measures being used are equivalent or functioning equally well across groups. Evidence of measurement equivalence is important for scientific inference and lack of measurement invariance casts suspicion and doubt on both conclusions drawn and on theory (Horn & McArdle, 1992). Although the importance of measurement equivalence recently has received increased attention and more research is being conducted examining the equivalence of various measures across groups, to date, there appears to be no prior research that has examined measurement equivalence in the body image field. This has created a potential flaw in many of the conclusions drawn in this field that relate to cross-group differences. Establishing measurement equivalence involves a series of sequential levels or tests Body Image 102 with each level examining the invariance of a different parameter or set of parameters. As the purpose of the present study was to assess whether three body image measures, the MBSRQ, the ASI-R, and the BIQLI, ultimately could be used to make gender and age group comparisons, the levels of invariance that were examined were configural, metric, and scalar invariance. Configural invariance, or invariance of factor patterns, indicates that the concept is fundamentally similar across groups. When configural invariance has been shown for a measure, the measure may be used to assess the concept of interest in each group and to explore the basic meaning and structure of the construct. However, because this level of invariance alone does not guarantee that the concept is being measured equally well in each group, it is not recommended to make comparisons across groups and interpret differences because these differences may reflect differences in the relevance of items among groups rather than actual differences on the latent construct itself. If configural invariance is not supported'for a measure, it indicates that the groups conceptualize the latent concept differently or that the measurement may have been confounded by some extraneous variables such as problems with the data collection (Cheung, & Rensvold, 2002). Metric invariance, or invariance of factor loadings, indicates that individuals from different groups interpret and respond to the measure in a similar way. When metric invariance has been shown for a measure, the measure may be used to examine structural relationships or correlations between the construct of interest and other constructs across groups (Steenkamp & Baumgartner, 1998). For instance, i f a researcher has a measure of body satisfaction that has been shown to be metrically invariant, the researcher may use this measure to examine whether the correlation between body satisfaction and self-esteem is higher for women than it is for men. Because this level of invariance cannot rule out whether there is measurement bias that exists that is causing the groups to respond differently to the items, it is not recommended Body Image 103 to make and interpret cross-group differences. If metric invariance is not supported for a measure, it could suggest that the latent construct is poorly operationalized or that cross-group differences exist in how the latent constrcut is conceptualized (Cheung, & Rensvold, 2000). Scalar invariance, or invariance of item intercepts, indicates that individuals who have the same value on the concept would obtain the same value on the observed variable regardless of their group membership. When scalar invariance has been shown for a measure, the measure may be used to assess cross-group differences on the observed scores on the measure. If scalar invariance is not supported for a measure, it indicates that bias exists in how the groups respond to the indicators. Two possible causes of this could be group differences in (a) levels of extreme response styles, whereby one group has the tendency to select the extreme points on a Likert-type scale, or (b) acquiescence response styles, whereby one group has a tendency to systematically give higher or lower responses (e.g., females always marking 2 points higher than males on a Likert-type scale; Cheung & Rensvold, 2000). A third possible cause relates to the relevance of the items defining the construct, whereby an item may be endorsed at a much higher rate for one group than the other because it "is more salient as a marker" of the construct for that group (Chan, 2000, p. 177). Multidimensional Body-Self Relations Questionnaire The results of the measurement invariance tests for the subscales of the M BSRQ clearly illustrate that the multidimensional nature of body image is perceived quite differently across the age and gender groups, as evidenced by no two subscales demonstrating the same level of invariance to the same degree. Table 30 provides a summary of the levels of invariance achieved for each subscale. Based on the varied findings for these subscales, it is important for researchers to consider the goals of their research when deciding which of these subscales to use. As the A E , A O , FO, IO, and OP subscales met the hypothesis of configural Body Image 104 invariance for the full model, the results support the use of these five subscales to assess their respective constructs across age and gender. It should be noted that the present study does not provide evidence that the construct being measured by each of these subscales is the construct that it claims to measure; rather, these results only suggest that the same construct is being assessed in each group. However, previous validation research with the various subscales of the MBSRQ has provided evidence to support the construct validity of this scale (Cash, 2000). Therefore, as an example, i f a researcher wanted to use the A E subscale as a measure of appearance evaluation in a sample of older women, the results of the present study support the use of this subscale for that purpose. Table 30 Levels of Invariance Attained for the Subscales of the MBSRQ Subscale Configural Metric Scalar Appearance Evaluation Full model Appearance Orientation Full model Female model Female model Young adult model Full model Middle-aged adult model Fitness Orientation Full model Full model Male model Young adult model Middle-aged adult model Older adult model Health Evaluation Male model Older adult model Health Orientation Male model Older adult model Male model Illness Orientation Full model Female model Young adult model Older adult model Young adult model Body Image 105 Body Areas Satisfaction Young adult model Young adult model Overweight Preoccupation Full model Full model Young adult model Middle-aged adult model Older adult model For the HE and HO subscales, configural invariance was supported for males and for older adults, suggesting that these subscales may be used to assess these constructs for adult men of all ages and for older adults. The lack of configural invariance for females, young adults, and middle-aged adults suggests that the constructs of H E and HO are not fundamentally similar across these groups in these models and thus these subscales should not be used to assess these constructs across women of all ages, with samples of both young men and women, or with samples of both middle-aged men and women. However, as the PCAs for the HE subscale only indicated the presence of one factor for each group of women, I propose that this subscale may be used with these groups to assess the construct of HE, but researchers should use some caution in interpreting the findings as there may be one or more minor secondary factor that may have a slight impact on the results. As the PCAs for the HO subscale indicated the presence of two factors for the young and middle-aged women, I do not recommend that this subscale be used with these two groups of women to assess investment in a healthy lifestyle. For the BAS subscale, which also tends to be one of the more popular subscales of the . MBSRQ, the lack of configural invariance for all subgroups except for young adults should serve to caution researchers wanting to use this subscale to assess the construct of body areas satisfaction across age and gender, as it appears that this construct is not fundamentally similar across these groups. For young men and women, the same unidimensional construct was found to be present in both groups and thus this subscale may be used to assess this construct across Body Image 106 gender with this age group. For the remaining groups, the results of the invariance testing suggest that this subscale should not be used in its present form. However, as the results of PCAs indicated the presence of one factor for all groups except for the middle-aged women, I propose that this subscale may be used with men of all ages and for young and older adult women, with the caveat that caution be used in interpreting the findings for middle-aged and older adult men and older adult women. I do not recommend that this subscale be used with middle-aged women to assess body areas satisfaction. Only three of the subscales (i.e., A O , FO, and OP) met the hypothesis of metric invariance for the full model. Thus, for these subscales, the results provide support for the use of these subscales to examine structural relationships, or correlations, between these subscales and other constructs across all age and gender groups. Four of the subscales (i.e., A E , HO, IO, BAS) showed support for metric invariance for one or more of the subgroups. For these subscales, it is only recommended to examine relationships among correlations for those subgroups meeting the hypothesis of metric invariance. That is, for the A E subscale, correlations can be examined for women and for young adults. For the HO subscale, correlations can be examined for men. For the IO subscale, correlations can be examined for women, young adults, and older adults. For the BAS subscale, correlations can be examined for young adults. For the remaining subgroups, and for the HE subscale, it is not recommended to examine correlations across groups. ' None of the subscales met the hypothesis of scalar invariance for the full model. Thus, these results suggest that researchers should not be using any of these subscales in their present forms to make comparisons across the six age and gender groups together. Five of the subscales (i.e., A E , A O , FO, IO, OP), however, did meet scalar invariance requirements for at least one or more of the subgroups. For the A E subscale, the results of the present study Body Image 107 indicated that comparisons may be conducted across age groups for women. For the A O subscale, gender comparisons may be conducted across middle-aged adults. For the FO subscale, comparisons may be conducted across all three age groups for men and across gender for all three age groups. For the IO subscale, gender comparisons may be conducted across young adults. Finally, for the OP subscale, gender comparisons may be conducted across all three age groups. For those subgroups and subscales that did not meet the requirements for scalar invariance, differences in item interpretation or measurement bias prevent accurate interpretation of observed mean differences. Examination of gender and age differences for those subscales exhibiting scalar invariance was fairly consistent with previous research using these subscales. The findings (a) that younger women report being more satisfied with their appearance, and (b) of a lack of difference in appearance satisfaction between middle-aged and older women, is in accordance with research by Cash and Henry (1995) and Deeks and McCabe (2001), but is contrary to research by Paxton and Phythian (1999) who found that middle-aged women reported greater appearance satisfaction than older women. The finding that middle-aged women invested more in their appearance than middle-aged men is consistent with the findings of Paxton and Phythian (1999). The finding that young adult men are more invested in their fitness level than young adult women is in accordance with research by Miller, Gleaves, Hirsch, Green, Snow, and Corbett (2000). This is not surprising given that men are more concerned with a muscular appearance as opposed to a thinner appearance (McCreary & Sasse, 2000) as evidenced by 91% of the young men in Jacobi and Cash's (1994) study reporting wanting to be more muscular. The finding that younger men reported higher fitness investment than middle-aged and older men may reflect more pressure that is beginning to be placed on younger men to value the male muscular body (Pope, Phillips, & Olivardia, 2000). The finding that young Body Image 108 adult women reported a greater awareness of physical symptoms and an increased likelihood of seeking medical attention is consistent with the findings of Miller et al. (2000). Finally, the finding that women scored higher than men on overweight preoccupation is also consistent with the research of Miller et al. (2000). Appearance Schemas Inventory-Revised For the ASI-R, the full two-factor model met the hypothesis of configural invariance, supporting that this scale is composed of two-factors. The demonstration of configural invariance also supports the use of this scale to assess these two factors across age and gender. Based on validation evidence by Cash et al. (2004), these two factors are said to measure the constructs of Self-Evaluative Salience and Motivational Salience. Therefore, as an example, i f a researcher was studying body image in a group of men, the results of the present study support the use of this scale to assess SES and MS in this group. The failure of the full model to meet the hypothesis of metric invariance indicates that there are cross-group differences in how participants respond to and interpret the items. Follow-up analyses for the five age and gender subgroups indicated that metric invariance was established for the male model, young adult model, and middle-aged adult model. These results provide support for the use of these two subscales to examine correlations for men, young adults, and middle-aged adults. Based on the overall findings for this scale, an additional post hoc analysis was conducted examining whether metric invariance could be established for the young and middle-aged females. The attainment of metric invariance for these two groups of women indicates that correlations may also be examined for young and middle-aged women. It also suggests that it is the group of older women who do not perceive the constructs of SES and MS in the same manner as the other groups. Possible reasons for this may be that the older women do not interpret the items in the same way or that the items may Body Image 109 not be as relevant to these women as they are in the remaining groups. As the construct is not being measured equally well for the older women, it is not advisable to examine correlations using these subscales with this age group. The results of the scalar invariance tests for the male model, young adult model, middle-aged adult model, and reduced female model (young and middle-aged women) indicated that comparisons may be conducted across age for men, across gender for young and middle-aged adults, and across age for young and middle-aged women. The only group for which comparisons should not be made is for the older women. As Cash et al. (2004) also propose that a Composite score can be calculated in addition to the two subscale scores, invariance testing was also examined for a one-factor model. Configural invariance for the one-factor model failed to be met for the full model and for all five gender and age subgroups, suggesting that the scale is not unidimensional and that a total score should not be used. Given that the multidimensional nature of the ASI-R is already recognized by its author through the SES and MS subscales, and that the results of the invariance testing support a two-factor model, to treat these subscales as unidimensional through the use of a composite score results in a loss of important information provided by the scale's multidimensionality. In addition, the interpretation of the composite score may become confusing as findings aggregate and may result in interpretational ambiguities. For example, an external criterion that correlates positively with only one of the underlying dimensions will demonstrate a lesser positive correlation with the aggregated score. Thus, based on the results of the present study, I do not recommend the use of a Composite score for the ASI-R. Researchers wishing to use this scale should analyze their data separately for the two subscales. Examination of gender differences for the two subscales indicated that both young and Body Image 110 middle-aged women reported significantly more self-evaluative investment, and greater levels of motivational salience on the ASI-R than their male counterparts. These findings are consistent with the research by Cash et al. (2004) who also found that young women reported higher scores than men. Examination of age differences for the SES subscale indicated that older men reported significantly less self-evaluative investment than both young and middle-aged men and young women reported significantly more self-evaluative salience investment than middle-aged women. These findings are consistent with the research by Pliner et al. (1990) and Tiggemann and Lynch (2001) who found that appearance importance tends to decrease with age. For the MS subscale, there were no significant age differences among men in all three age groups or among young and middle-aged women. These results suggest that age does not appear to be a factor in how much individuals engage in appearance-management behaviors. Body Image Quality of Life Inventory For the BIQLI, missing data was particularly problematic for item 6, "My experiences at work or school", item 11, "My feelings of acceptability as a sexual partner", and item 12, "My enjoyment of my sex life". In particular, approximately 10% of the oldest group of women had missing responses for one, two, or all three of these items. Although the percentage of missing data for the oldest group of men was not nearly as high as that of the older women, of the missing responses for this age group, half of them were from items 6, 11, and 12. On a few of the paper questionnaires, participants, both men and women, indicated they did not respond to these items because these items were no longer applicable to them. Item 12 may also be problematic for those individuals who have not had sex, or do not have an active sex life. For this reason, missing data was not imputed for these items. Separate analyses were conducted with and without these items to investigate whether the absence of Body Image 111 these three items would impact the conclusions drawn about the scale. Both the full scale and reduced scale produced similar results, indicating that the removal of these items did not impact the factor structure or the degree of measurement invariance attained across the models tested. Thus, the results of the measurement invariance tests will be discussed in terms of the full model only. For researchers wishing to use this scale, it is recommended that the items should be retained in the scale, as they are highly relevant to body image quality of life. For those individuals with missing responses for any of these items, a composite score may still be calculated by taking the mean of the number of items that were answered. A l l six age and gender groups met the hypotheses of configural, metric, and scalar invariance for this measure. These results indicate that this scale can be used to make comparisons of observed scores across gender and age. Given that the BIQLI is a recently developed scale that has undergone little empirical research, these findings provide initial support for the use of this measure for examining gender and age differences. Examination of gender and age differences revealed that men reported a significantly more favorable body image quality of life than did women, and that older adults reported a significantly more favorable body image quality of life than did younger adults, who in turn reported a significantly more favorable body image quality of life than did middle-aged adults. The finding of a significant gender difference is in accord with the findings of Cash, Jakatdar, and Williams' (2003) study using college males and females. Conclusion There are three important implications of this research that should be noted. One, the findings for all three scales, and in particular the MBSRQ, highlight the importance of examining measurement equivalence for a measure before that measure is used to assess the construct of interest across different groups. Failure to do so may result in researchers Body Image 112 measuring a construct they are not intending to. This, in turn, can impact the validity of conclusions drawn and may distort ensuing theory. This does not mean that a single study of measurement invariance for a measure is adequate to conclude that the scale is invariant for all future uses. Ideally, a researcher should examine measurement invariance each and every time a measure is used in a new sample. However, in all practicality, this may not be possible for all researchers due to time constraints and/or difficulty of analyses. I advocate that, i f measurement invariance has been demonstrated repeatedly for a measure over varied samples and time periods, a researcher can gain confidence that the measure will be invariant for those groups examined without having to check for measurement invariance himself or herself. That said, as the present study is the first study to examine the measurement invariance of the MBSRQ, the ASI-R, and the BIQLI, these findings should be interpreted as showing initial support for the invariance, or lack of invariance, of these measures across age and gender, and that more research is needed to replicate these findings. Researchers interested in conducting measurement invariance analyses using LISREL are referred to Vandenberg and Lance (2000) and a website guide entitled "Studying Measurement Invariance" (n.d.). Two, these findings should also serve as a warning to researchers that caution should be used in deciding what scales/subscales are the most appropriate for the purposes of their research. That is, i f researchers are interested in examining gender differences, they should only use those scales/subscales that have been found to show scalar invariance for men and women, such as the FO subscale of the MBSRQ, or the BIQLI. If researchers are interested in examining relationships between body image and other measures of self-concept, then they need to choose scales/subscales that have been found to be metrically invariant for the groups they are interested in, such as the A O subscale of the MBSRQ, or the ASI-R. Finally, i f researchers are only looking for a scale to measure a specific construct for a particular group of Body Image 113 individuals, such as young men and women, they only need to choose a scale/subscale that has demonstrated configural invariance for those groups, such as the B A S subscale of the MBSRQ. Three, these results must also call into question the findings of any cross-group differences or correlational research that have been made in the past. For instance, Paxton and Phythian (1999) used the MBSRQ in a sample of middle-aged and older adult men and women to examine gender and age effects. Several of their conclusions on gender differences are based on MBSRQ subscales, such as Appearance Orientation and Health Orientation, which have not been shown in this study to exhibit scalar invariance for these groups. Thus, the findings that women scored higher than men may likely be the result of differences in interpretation and measurement bias on the items for this subscale. However, that is not to say that the gender differences do not necessarily exist, but rather that any such differences may not be an accurate assessment of the true degree of difference. By claiming gender differences on these subscales when it is not appropriate to do so, the theory to which these findings contribute may become distorted or erroneous. As both the ASI-R and BIQLI are newly developed scales, very limited research has been conducted with these scales thus far. Future research with the scales-must take into consideration the characteristics of the sample and the goals of the study. Strengths of the Study There are four major strengths of the present research that distinguish it from previous research in the body image field. First, the present research is the only study to have examined measurement invariance for the MBSRQ, ASI-R, and BIQLI, and, to my knowledge, is the first study to examine measurement invariance in the body image literature. Given the differing findings for the various scales/subscales in the present study, this research highlights Body Image 114 the importance of examining measurement invariance when conducting body image research and can hopefully serve as a springboard to encourage the continued investigation of measurement invariance in this field. Second, middle-aged and older adults were included in the present research. The majority of previous research on body image has generally focused on adolescents and young adult college students. Thus, the present study has extended the literature into these understudied populations through the inclusion of a community-based sample of men and women covering a wider age range than the majority of previous studies on body image. Third, the large sample size used in the present research is a strength. This research obtained a much larger sample size than what is typically seen in the body image field. By including a large number of participants, I was able to examine measurement invariance across both age and gender. Finally, there are also several strengths associated with the use of a web-based survey in addition to the more traditional paper-and-pencil survey. By using a web survey, I was able to collect data from a much larger sample size in a much shorter time than I would have been able to through paper-and-pencil methods alone. The web-based survey was also much more convenient for participants, allowing them to complete the survey at a time and location of their choice. Finally, a web-based survey generated a more diverse sample by allowing me to access participants all over Canada and the US. Limitations There are three limitations in the present research that I will discuss. First, I believe that the length of the questionnaire, which took approximately 30 minutes to complete, may have been too long for some participants. Had the questionnaire been about 10 to 15 minutes in length, more participants may have been willing to take part and to complete the entire Body Image 115 survey. Questionnaire length may have had the biggest impact on the older adults and may partially explain why I had great difficulty in getting the numbers I needed for this age group. Second, the topic of body image may have been perceived by a lot of men as being more of a female topic. I believe this may explain why it was more difficult to get men, especially middle-aged and older men, to take part in the survey. Because male participants were more difficult to recruit, the time limitation was a factor in being able to recruit sufficient numbers of men in all age groups. Although I was still able to conduct all planned analyses, the smaller sample sizes for these groups (i.e., less than the original goal of 200) may make the results obtained for these groups a little less reliable. Finally, as this is the first study of its kind, the findings of this research need to be replicated. In particular, some of the decisions for configural invariance were based on fit statistics that just met or did not meet the cutoff values for indicating invariance (i.e., a CFI value of .89 was interpreted as not indicating invariance, but a CFI value of .90 was said to support invariance). If the research were to be conducted again, there is a chance that some of the borderline results may swing the other way. Thus, more research is needed to verify these findings. Future Directions I will discuss four directions for future research. First, as mentioned previously, because this is the first study examining measurement invariance for the MBSRQ, ASI-R, and BIQLI, future research is needed to replicate these findings. In addition, as evidenced by the varying findings for the three scales examined in the present study, this research should highlight the importance of examining measurement invariance for all body image measures. Second, more research is needed to examine why various levels of invariance have not been met for the subscales of the MBSRQ and for the older women on the ASI-R. For Body Image 116 instance, it really should not be too surprising that scalar invariance, or even metric or configural invariance, was not met for men and women across the adult age range for the HE or HO subscales given that health concerns become much more important to one's body image than appearance as individuals get older (Clarke, 2002). However, the lack of invariance for other subscales is not as easy to interpret. More research is needed to investigate possible sources of invariance and ways to potentially overcome this i f one wishes to make cross-group comparisons. Third, another area of research could focus on the issue of partial measurement invariance. The present study only focused on whether the three levels of invariance were fully satisfied or not. However, Byrne, Shavelson, and Muthen (1989) recommend the testing of partial measurement invariance. Thus, future research could examine whether partial measurement invariance can be achieved for all or some of these subscales. Those items meeting the requirements for partial invariance could then be compared across groups, or those items not meeting invariance could be revised to be more appropriate for all groups. Finally, more research is needed to examine the relationship between configural invariance and factor structure. In the present study, there were a couple of examples where configural invariance was not met for the full model. However, follow-up factor analyses suggested the same factor structure (i.e., one-factor) in each subgroup that was specified in the configural invariance model. Research is needed to explore possible reasons for the lack of configural invariance when all groups show a unidimensional factor structure. Body Image 117 References Babakus, E., Ferguson, C. E., Joreskog, K. G. (1987). The sensitivity of confirmatory maximum likelihood factor analysis to violations of measurement scale and distributional assumptions. Journal of Marketing Research, 24, 222-228. Baumgartner & Steenkamp (2004). The survey research/theory testing paradigm. Retrieved May 1, 2005, from http://www.marketingadvances.org/confprior/2004Steenkamp.ppt Benson, J., & Fleishman, J. (1994). The robustness of maximum likelihood and distribution-free estimators to non-normality in confirmatory factor analysis. Quality & Quantity, 28, 117-136. Bentler, P. M . (1990). Comparative fit indexes in structural models. Psychological Bulletin, 107, 238-246. Bentler, P. M . (1998). Causal modeling: New interfaces and new statistics. In J. G. Adair & D. Belanger (Eds.), Advances in psychological science, vol. 1: Social, personal, and cultural aspects. Hove, England: Erlbaum. Bentler, P. M . , & Bonnett, D. G. (1980). Significance tests and goodness-of-fit in the analysis of covariance structures. Psychological Bulletin, 88, 588-606. Bollen, K. A . , & Long, S. J. (1993). Testing structural equation models. Newbury Park, C A : Sage Publications. Bozdogan, H . (1987). Model selection and Akaike's information criterion (AIC): The general theory and its analytical extensions. Psychometrika, 52, 345-370. Brannick, M . T. (1995). Critical comment on applying covariance structure modeling. Journal of Organizational Behavior, 16, 201-213. Browne, M . W. (1984). Asymptotic distribution free methods in the analysis of covariance structures. British Journal of Mathematical and Statistical Psychology, 37, 127-141. Body Image 118 Browne, M . W., & Cudeck, R. (1989). Alternative ways of assessing model fit. In K . A . Bollen, & J. S. Long (Eds.), Testing structural equation models (pp. 136-162). Newbury Park, C A : Sage Publications. Byrne, B. M . , Shavelson, R. J., & Muthen, B. (1989). Testing for the equivalence of factor covariance and mean structures: The issue of partial measurement invariance. Psychological Bulletin, 105, 456-466. Cash, T. F. (2000). The Multidimensional Body-Self Relations Questionnaire users' manual. Available from the author at www.body-images.com. Cash, T. F. (2003). Brief manual for the Appearance Schemas Inventory-Revised. Available from the author at www.body-images.com. Cash, T. F., & Fleming, E. C. (2002). The impact of body image experiences: Development of the Body Image Quality of Life Inventory. International Journal of Eating Disorders, 31, 455-460. Cash, T. F., & Henry, P. E. (1995). Women's body images: The results of a national survey in the U.S.A. Sex Roles, 33, 19-28. Cash, T. F., Jakatdar, T. A. , & Williams, E. F. (2003). The Body Image Quality of Life Inventory: Further validation with college men and women. Body Image, 1, 279-287. Cash, T. F., Melnyk, S. E., & Hrabosky, J. I. (2004). The assessment of body-image assessment: A n extensive revision of the Appearance Schemas Inventory. International Journal of Eating Disorders, 35, 305-316. Cash, T. F., & Pruzinsky, T. (Eds.). (1990). Body images: Development, deviance, and change. New York: Guildford Press. Cash, T. F., Winstead, B. A . , & Janda, J. H . (1986). Body image survey report: The great American shape-up. Psychology Today, April, 30-37. Body Image 119 Chan, D. (2000). Detection of differential item functioning on the Kirton Adaptation-Innovation Inventory using multiple-group mean and covariance structure analyses. Multivariate Behavioral Research, 35, 169-199. Cheung, G. W., & Rensvold, R. B. (2000). Assessing extreme and acquiescence response sets in cross-cultural research using structural equation modeling. Journal of Cross Cultural Psychology, 31, 187-212. Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling, 9, 233-255. Chrisler, J. C , & Ghiz L. G. (1993). Body image issues of older women. Special issue: Faces of women aging. Women and Therapy, 14, 67-75. Clarke, L. H . (2002). Older women's bodies and the self: The construction of identity in later life. Canadian Review of Sociology & Anthropology, 38, 441-464. Cudeck, R. (1989). Analysis of correlation matrices for covariance structure models. Psychological Bulletin, 105, 317-327. Curran, P. J., West, S. G., & Finch, J. F. (1996). The robustness of test statistics to nonnormality and specification error in confirmatory factor analysis. Psychological Methods, 1, 16-29. Deeks, A . A. , & McCabe, M . P. (2001). Menopausal stage and age and perceptions of body image. Psychology and Health, 16, 367-379. Diamantopoulos, A . , & Siguaw, J. A . (2000). Introducing LISREL: A guide for the uninitiated. London, England: Sage. Dolan, C. V . (1994). Factor analysis of variables with 2, 3, 5, and 7 response categories: A comparison of categorical variable estimators using simulated data. British Journal of Mathematical and Statistical Psychology, 47, 309-326. Body Image 120 Fabrigar, L. R., Wegener, D. T., MacCallum, R. C , & Strahan, E. J (1999). Evaluating the use of exploratory factor analysis in psychological research. Psychological Methods, 4, 272-299. Fallon, A . E., & Rozin, P. (1985). Sex differences in perceptions of desirable body shape. Journal of Abnormal Psychology, 94, 102-105. Garner, D. M . (1997). The 1997 body image survey results. Psychology Today, 30, 30-36, 38-40, 42-44, 75-76, 78, 84. Gosling, S. D., Vazire, S., Srivastava, S., & John, O. P. (2004). Should we trust web-based studies? A comparative analysis of six preconceptions about internet questionnaires. American Psychologist, 59, 93-104. Gupta, M . A. , Schork, N . J. (1993). Aging-related concerns and body image: Possible future implications for eating disorders. International Journal of Eating Disorders, 14, 481-486. Horn, J. L. (1991). Discussion of the issues of factorial invariance. In L. M . Collins & J. L . Horn (Eds.), Best methods for the analysis of change (pp. 114-125). Washington, DC: American Psychological Association. Horn, J. L., & McArdle, J. J. (1992). A practical and theoretical guide to measurement invariance in aging research. Journal of Experimental Aging Research, 18, 117-144. Horn, J. L. , McArdle, J. J., & Mason, R. (1983). When is invariance not invariant: A practical scientist's look at the ethereal concept of factor invariance. The Southern Psychologist, 1, 179-188. Hox, J. J. (1999). A review of current software for handling missing data. Kwantitatieve Methoden, 62, 123-138. Hu. L.T., & Bentler, P. M . (1993). Evaluating model fit. In R. H . Hoyle (Ed.), Structural Body Image 121 equation modeling: Concepts, issues, and applications (pp. 76-99). Thousand Oaks, C A : Sage Publications. Hu, L. T., & Bentler, P. M . (1998). Fit indices in covariance structure modeling: Sensitivity to underparameterized model misspecification. Psychological Methods, 3, 424-453. Hu, L. T., & Bentler, P. M . (1999). Cutoff criteria for fit indices in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55. Hurd, L . C. (2000). Older women's body image and embodied experience: A n exploration. Journal of Women & Aging, 12(3/4), 77-97. Jacobi, L. , & Cash, T. F. (1994). In pursuit of the perfect appearance: Discrepancies among self-ideal percepts of multiple physical attributes. Journal of Applied Social Psychology, 24, 379-396. Janelli, L. M . (1993). Are there body image differences between older men and women? Western Journal of Nursing Research, 15, 327-339. Joreskog, K . G. (1971). Simultaneous factor analysis in several populations. Psychometrika, 36, 409-426. Joreskog, K . G., & Sorbom, D. (1993). LISREL 8: User's reference.guide. Chicago, IL: Scientific Software International. Kaiser, H . F. (1970). A second generation Little Jiffy. Psychometrika, 35, 401-415. Kelloway, K . E. (1995). Structural equation modelling in perspective. Journal of Organizational Behaviour, 16, 215-224. Kirk, R. E. (1996). Practical significance: A concept whose time has come. Educational & Psychological Measurement, 56, 746-759. Body Image 122 Kline, R. B. (1998). Principles and practice of structural equation modeling. New York: Guilford Press. Kreuger, D. W. (1989). Body self and psychological self. New York: Brunner/Mazel. Lewis, D. M . , & Cachelin, F. M . (2001). Body image, dissatisfaction, and eating attitudes in midlife and elderly women. Eating Disorders, 9, 29-39. Lokken, K. , Ferraro, R. F., Kirchner, T., & Bowling, M . (2003). Gender differences in body size dissatisfaction among individuals low, medium, or high levels of body focus. Journal of General Psychology, 130, 305-310. Marshall, V . W. (1983). Generations, age groups and cohorts: Conceptual distinctions. Canadian Journal of Aging, 2, 51-61. McCreary, D. R., & Sasse, D. K . (2000). An exploration of the drive for muscularity in adolescent boys and girls. Journal of American College Health, 48, 297-304. Meredith, W. (1993). Measurement invariance, factor analysis and factorial invariance. Psychometrika, 58, 525-543. Meredith, W. (1995). Two wrongs may not make a right. Multivariate Behavioral Research, 30, 89-94. Miller, K. J., Gleaves, D. H. , Hirsch, T. G., Green, B. A. , Snow, A . C , & Corbett, C. C. (2000). Comparisons of body image dimensions by race/ethnicity and gender in a university population. International Journal of Eating Disorders, 27, 310-316. Muth, J. L. , & Cash, T. F. (1997). Body-image attitudes: What differences does gender make? Journal of Applied Social Psychology, 27, 1438-1452. Muthen, B., & Kaplan, D. (1985). A comparison of some methodologies for the factor analysis of non-normal Likert variables. British Journal of Mathematical and Statistical Psychology, 38, 171-189. Body Image 123 Muthen, B., & Kaplan, D. (1992). A comparison of some methodologies for the factor analysis of non-normal Likert variables: A note on the size of the model. British Journal of Mathematical and Statistical Psychology, 45, 19-30. Novak, M . (1997). Aging and society: A Canadian perspective. Toronto: ITP Nelson. Nowak, C. A. , (1977). Does youthfulness equal attractiveness? In L. E. Troll, J. Israel, & K . Israel (Eds.), Looking ahead: A woman's guide to the problems and joys of growing older (pp. 59-64) Englewood Cliffs, NJ: Prentice-Hall. Oberg, P., & Tornstam, L. (2001). Youthfulness and fitness - Identity ideals for all ages? Journal of Aging and Identity, 6, 15-29. Olsson, U . (1979). Maximum likelihood estimation of the polychoric correlation coefficient. Psychometrika, 44, 443-460. Palmore, E. (1978). When can age, period, and cohort be separated? Social Forces, 57, 282-295. Paxton, S. J., & Phythian, K . (1999). Body image, self-esteem, and health status in middle and later adulthood. A ustralian Psychologist, J^, 116-121. Pliner, P., Chaiken, S., & Flett, G. L . (1990). Gender differences in concern with body weight and physical appearance over the life span. Personality and Social Psychology Bulletin, 16, 263-273. Pope, H . G., Phillips, K . A . , & Olivardia, R. (2000). The Adonis complex: The secret crisis of male body obsession. New York: Free Press. Reise, S. P., Waller, N . G., & Comrey, A . L. (2000). Factor analysis and scale revision. Psychological Assessment, 12, 287-297. Rigdon, E. E., Ferguson, C. E., Jr. (1991). The performance of the polychoric correlation coefficient and selected fitting functions in confirmatory factor analysis with ordinal Body Image 124 data. Journal ofMarketing Research, 28, 491-197. Rock, D. A. , Werts, C. E., & Flaugher, R. L . (1978). The use of analysis of covariance structures for comparing the psychometric properties of multiple variables across populations. Behavioral Research, 13, 403-418. Rozin, P., & Fallon, A . (1988). Body image, attitudes to weight, and misperceptions of figure preferences of the opposite sex: A comparison of men and women in two generations. Journal of Abnormal Psychology, 97, 342-345. Schafer, J. L. , Olsen, M . K . (1998). Multiple imputation for multivariate missing-data problems: A data analyst's perspective. Multivariate Behavioral Research, 33, 545-571. Slade, P. D. (1988). Body image in anorexia nervosa. Journal of Psychiatry, 153, 20-22. Sontag, S. (1979). The double standard of aging. In J. Williams (Ed.), Psychology of women (pp. 462-478). San Diego, C A : Academic Press. Stanford, J. N . , & McCabe, M . P. (2002). Body image ideal among males and females: Sociocultural influences and focus on different body parts. Journal of Health Psychology, 7, 675-684. Statistics Canada. (2004). Population by sex and age group. Retrieved May 20, 2005, from http://www40.statcan.ca/101 /cstO 1 /demo 10a.htm?sdi=age Steenkamp, J-B. E. M . , & Baumgartner, H . (1998). Assessing measurement invariance in cross-national consumer research. Journal of Consumer Research, 25, 78-90. Steiger, J. H . (1990). Structural model evaluation and modification: A n interval estimation approach. Multivariate Behavioral Research, 25, 173-180. Studying measurement invariance using confirmatory factor analysis, (n.d.). Retrieved August 24, 2005, from http://www.unc.edu/~rcm/psy236/measinv.pdf Body Image 125 Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics, 4 t h ed. Boston: Allyn and Bacon. Tiggemann, M . , & Lynch, J. E. (2001). Body'image across the life span in adult women: The role of self-objectification. Developmental Psychology, 37, 243-253. Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis of the measurement invariance literature: Suggestions, practices, and recommendations for organizational research. Organizational Research Methods, 3, 4-69. Webster, J., Tiggemann, M . (2003). The relationship between women's body satisfaction and self-image across the life span: The role of cognitive control. The Journal of GeneticPsychology, 164, 241-252. Wilcox, S. (1997). Age and gender in relation to body attitudes. Psychology of Women Quarterly, 21, 549-565. Wolf, N . (1991). The beauty myth. How images of beauty are used against women. New York: Morrow. Yoo, B. (2002). Cross-group comparisons: A cautionary note. Psychology & Marketing, 19, 357-368. Ziebland, S., Robertson, J., Jay, J., & Neil, A . (2002). Body image and weight change in middle age: A qualitative study. International Journal of Obesity, 26, 1083-1091. Zumbo, B. D. & Hubley, A . M . (1998). A note on misconceptions concerning prospective and retrospective power. Journal of the Royal Statistical Society, Series D: The Statistician, 47, 385-388. Body Image 126 Appendix A Personal Demographic Information Please answer the following demographics questions as accurately as you can. A l l the information that you provide on this form is confidential. This information is collected solely for the purpose of describing the sample. 1. Gender: D Male D Female 2. Age: 3. What is your primary ethnic/racial/cultural background? (please check one box only) D Aboriginal or First Nations D East Asian (e.g., Chinese, Japanese, Korean) D Hispanic (e.g., Latino, Mexican) D South Asian / Middle Eastern (e.g., Indian, Arabic) D White (e.g., Caucasian, Anglo, European Origin) D Other: (please specify: ) (note: Please do not record nationality, such as "Canadian", for this question.) 4. Were you born in Canada or the US? D no D yes If no, how long have you lived in Canada or the US? 5. What is the highest level of education that you have completed? D Less than high school D High school graduate D Some college or university D Diploma or Associate's Degree (1 or 2 year program) D Bachelor's Degree (3 or 4 year degree) D Graduate study 6. What is your present marital status? D Married or common-law • Widowed D Divorced/Separated D Never Married Body Image 127 7. Please indicate if you have been clinically diagnosed with any of the following conditions / within the last 3 years. (Please check as many as apply.) • Anorexia • Bulimia • Other eating disorder (please specify • HIV/AIDS • Cancer • None of the above 8. Do you have any physical deformities, conditions, or imperfections (e.g., burns or scars, missing limbs, birthmarks or skin conditions, weight or height issues, etc.) that you feel impact your body image? • no • yes (please specify ) 9. Do you have any physical deformities, conditions, or imperfections (e.g., burns or scars, missing limbs, birthmarks or skin conditions, weight or height issues, etc.) that others might expect to impact your body image? • no • yes (please specify ) 10. Height: ' 11. Weight (in pounds): Questions 12-14 are for women only 12. Are you pregnant? • no • yes 13. Have you been pregnant or given birth in the last year? D no D yes 14. Have you started or gone through menopause? • no • yes 

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}]}"
                            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.831.1-0054394/manifest

Comment

Related Items