Growth Models for Categorical Response Variables: Standard, Latent-Class, and Hybrid Approaches

This chapter deals with growth models for repeatedly observed categorical response variables, there are other alternatives for analyzing the type of data. Variants of transitional, growth, and marginal models have been developed for categorical response variables, as well as hybrids combining the approaches. The chapter describes the standard random-effects growth model for categorical response variables. It discusses the latent-class (LC) based growth modeling approach, as well as hybrid variants combining discrete and continuous random effects. The chapter presents an empirical example in which the standard, LC, and hybrid methods are applied. It compares the obtained results with the ones that would have been obtained with a Markov model. The chapter also presents extensions of the basic models that involve the use of methods for three-level instead of two-level data. Some of the problems associated with the parametric random-coefficients approach may be circumvented by adopting a latent-class based nonparametric random coefficients approach.