Modeling interactions in latent transition analysis (LTA)

In recent years the latent class analysis (LCA) has been increasingly applied in the fields of public health studies. However, applications of the latent transition analysis (LTA), which is an extension of LCA to longitudinal data, are still very limited. This is partially because of the difficulty in modeling the impact of a covariate (e.g., intervention/treatment) on latent class transitions that involves interactions between the covariate and latent class. Using simulated data, this study demonstrates how to use Mplus to estimate LTA model with a 3x3 transition matrix (i.e., three classes in the latent class variables at both time 1 and time 2, each of which has four dichotomous indicators). The structural parameters of the latent transitions will be estimated using different parameterization approaches. Our results provide evidence that different parameterization approaches for modeling interactions between covariate and latent class in LTA model provide equivalent results. Parameters estimated from both Parameterizations 1 and 2 can be used to assess the effects of covariates on latent transitions in LTA model.