The 131st Annual Meeting (November 15-19, 2003) of APHA

Abstract #53122

Jichuan Wang, PhD, Harvey Siegal, PhD, Russel Falck, MA, and Robert Carlson, PhD. Community Health, Wright State University, 3640 Colonel Glenn Hwy, Dayton, OH 45435, 937-775-2066, jichuan.wang@wright.edu

The recently developed growth mixture modeling (GMM), which combines the latent growth model (LGM) and cluster analysis, has drawn increasingly attention in public health research using longitudinal data. The GMM not only allows the intra-individual changes in outcome measures to be analyzed together with inter-individual differences, it also captures the heterogeneity of growth by including a categorical latent variable (the latent class variable); therefore, allows different subclasses in the population to follow different growth trajectories. This study demonstrates how to apply the GMM to examine classes of growth trajectories of outcome measures. The Bayesian Information Criterion (BIC) and the approximation of Bayes factor are used to identify the optimal number of classes of trajectories. The average individual posterior probability for being assigned in each class, as well as the summary measure of classification (i.e., Entropy measure), are used to assess the classification quality. And the predictors of group membership probability are examined. In addition, the appropriate shape of growth trajectory within each class is tested; and background variables are used to predict the growth trajectory within each class. A sample of a three-year natural history study on health service utilizations among 430 crack-cocaine users is used for example modeling in this study.

Learning Objectives:

At the conclusion of the presentation, participants would be able to know:
• the usefulness of GMM for public health research;
• the fundamentals of GMM; 3)the appropriate computer software for GMM

Keywords: Statistics, Substance Abuse

Presenting author's disclosure statement:
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.