ABSTRACT
This chapter presents a procedure that accounts for sample heterogeneity—finite mixtures—and their application to longitudinal data. It briefly outlines the procedures for conducting latent variable growth mixture modeling. In growth mixture modeling, the modeling of repeated measures of the binary or ordered categorical latent class indicators also is possible. It was shown that growth mixture modeling can be used instead of the more traditional latent growth curve model to reduce dimensionality and estimate a separate regression model for each subpopulation (group). The implementation of random starting values increases the likelihood of convergence and of a unique model solution. The growth mixture modeling approach presented is strengthened by its association with the general latent variable modeling framework, and offers the potential for exploring new and more complex theories of development among a plethora of behaviors.
