ABSTRACT
This chapter discusses the role of random effect models for dealing with heterogeneity in life history processes more extensively, along with their role in modeling dependence between different multistate processes. It describes alternative approaches for dealing with dependence between multistate processes as are other applications of models involving latent processes. Random effect models can play a useful role in the analyses of multistate processes for similar reasons to the survival setting, but there are issues warranting careful consideration. Models accommodating heterogeneity through the use of continuous random effects involve very particular assumptions on the nature of heterogeneity and are hard to contemplate in non-progressive processes with many states. In semi-Markov models, the analysis is based on the sojourn time in states, and copula models naturally lend themselves to dependence modeling among sojourn times. The chapter considers two Markov models for describing the progression of retinopathy as measured by a 23-point ordinal scale such as Model M1B and M2B.
