ABSTRACT
This chapter presents methods for fitting multistate models for the analysis of data from individuals who are observed continuously over a period of time in a prospective study. It describes likelihoods and discusses methods for estimation and inference for parametric models. The chapter considers nonparametric methods and semiparametric analyses and provides examples to illustrate the use of software and the interpretation of results. In some settings it may be best to use Markov intensities for certain types of transitions and semi-Markov intensities for others. The chapter refers to these processes as having "hybrid" time scales and provides an illustration involving such a formulation where the authors model the onset and resolution of outbreaks of symptoms in individuals infected with the herpes simplex virus. The irreversible illness-death model is commonly used to characterize risk of tumour development and to assess the lethality of tumours in animal carcinogenicity experiments.
