Mixture Models for Multilevel Data Sets
The aim of this chapter is threefold. The first is to explain the relationship between LC analysis and multilevel regression analysis techniques. It will be shown that LC models can be conceptualized as models for two-level data sets in which parameters vary randomly across level-2 units. Whereas in multilevel regression analysis this variation is modeled by assuming that parameters come from a particular continuous distribution (typically normal), which is equivalent to introducing one or more continuous latent variables in the model, in LC analysis variation is modeled using discrete latent variables (Aitkin, 1999; Vermunt & Van Dijk, 2001).