Open Collections

NEXUS Spring Institute

Using latent variable mixture modelling to explore sample diversity Sawatzky, Rick; Richardson, Chris 2008-04-24

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

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

Item Metadata


60499-RichardsonSawtzky2008SpringInstituteApr24.pdf [ 680.65kB ]
JSON: 60499-1.0103699.json
JSON-LD: 60499-1.0103699-ld.json
RDF/XML (Pretty): 60499-1.0103699-rdf.xml
RDF/JSON: 60499-1.0103699-rdf.json
Turtle: 60499-1.0103699-turtle.txt
N-Triples: 60499-1.0103699-rdf-ntriples.txt
Original Record: 60499-1.0103699-source.json
Full Text

Full Text

Using Latent Variable Mixture Modeling to Explore Sample Diversity Dr. Rick Sawatzky, PhD, RN Nursing Department - Trinity Western University Dr. Chris G Richardson, PhD Research Scientist, CHÉOS Assistant Professor, Dept of Health Care and Epidemiology, UBC 2Emerging tobacco dependence • Tobacco use typically begins in adolescence • Emerging tobacco dependence is multidimensional • Diversity influences emergence and maintenance of dependence: Gender – Girls experience higher levels of nicotine dependence – Girls have a shorter time to withdrawal symptoms – Nicotine influences the subjective and reinforcing effects of tobacco use to a lesser degree in woman – Subjective and reinforcing effects of non-nicotinic stimuli associated with smoking may be greater in women 3Theoretical model for the DTDS Strong cravings Feel right Concentrate Fit in at school Look cool Feel popular Nicotine Smoking Sensory Emotional Social Stressed     Angry     Nervous Handling cigarettes     Taste     Blowing out smoke 4Diversity • Diversity is often meant to refer to known differences that are relevant to a particular analysis. – Demographic differences – Other relevant differences • The assumption is that one knows what the relevant differences are a priori. • We examined the plausibility of sources of diversity that are not known a priori by evaluating sample heterogeneity with respect to: a. The measurement structures of the DTDS subscales b. The profiles of the DTDS summary scores 5B.C. Youth Survey on Smoking and Health 2 • Cross-sectional survey conducted in 2004 • 8125 students • 1,425 smoked at least once in previous month • 55% female, average age of 16 years • Average of 4.1 cigs/day Methods • Factor Mixture Analysis • Latent Profile Analysis Data and methodology 6Factor mixture analysis An exploratory technique used to identify subgroups of respondents that differ with respect to the measurement of each of the DTDS dimensions. Research questions: 1. Are there subgroups of adolescents that respond to the DTDS questions in different ways? 2. Are these subgroups explained by demographic variables and other variables associated with tobacco dependence (e.g., life time # of cigarettes smoked)? 7Factor mixture model diagram Social ε d e p s 5 ε d e p s 7 ε d e p p w 3 ε d e p s 4 ε d e p s 2 ε d e p s 6 latent classes Ethnicity Gender Grade Depression Lifetime smoking 8FMAs of DTDS subscales Subscale K BIC Class proportions C1 C2 Nicotine 1 13621.91 1.00 2 13517.94 0.60 0.40 Emotion 1 43585.25 1.00 2 43162.00 0.60 0.40 Social 1 13629.07 1.00 2 13332.22 0.64 0.36 Sensory 1 15880.41 1.00 2 15559.85 0.65 0.35 K = number of latent classes. P = number of model parameters. BIC = Bayesian Information Criterion. 9Social dimension of the DTDS Smoking helps me feel good at school 10 Social dimension of the DTDS This item does not provided much information about adolescents in class 1. 11 Explaining latent classes 12 Explaining latent classes Explanatory variables Emotion (R2 = .01) Social (R2 = .01) Sensation (R2 = .03) Nicotine (R2 = .03) Gender Grade Ethnicity Depression # cigarettes smoked = Not significant = Significant (p < .05) 13 Explaining latent classes Explanatory variables Emotion Social Sensation Nicotine Gender nc Grade nc Ethnicity nc Depression nc # cigarettes smoked nc = Not significant = Significant (p < .01) 14 Factor mixture analysis results • Traditionally, look at differences on total scores each dimensions of the DTDS • But “diversity” leads to multiple pathways to addiction • LPA identifies subgroups of respondents with similar DTDS profiles 15 Latent profile analysis 16 DTDS profiles for the 3 classes 17 Correlations for the 3 classes Correlation Class 1 (n=255) Class 2 (n=391) Class 3 (n=476) Social with Nicotine .22 .17 .46 Emotion .03 .15 (NS) .37 Sensation .30 .55 .61 Nicotine with Emotion .46 .07 (NS) .72 Sensation .17 .17 .48 Emotion with Sensation .16 -.01 (NS) .40 18 Explaining latent profiles • Class structure should be informed by covariates • ALL relevant covariates should be included • Examine stability and fit of models as covariates included • More research needed on this methodology 19 Explaining latent classes Can indentify latent class membership BUT...


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items