ABSTRACT
In the previous application chapters we considered responses of particular types. In this chapter we exploit the generality of the general model framework and discuss applications where the responses are from multiple processes and possibly of mixed type. Combinations treated here are continuous and dichotomous, dichotomous and counts, continuous and continuous time survival and dichotomous and continuous time survival. Importantly, we will see that it is often not permissible to simply decompose such problems, that is by separately modeling the different processes. As in other application chapters, we will use continuous latent variables with parametric distributions and discrete latent variables, interpretable as latent classes or as a ‘nonparametric’ estimator of an unspecified distribution. The usefulness of structural models, regressing latent variables on other variables, will be demonstrated.
