Inference for non-markov multi-state models: an overview

Abstract

In longitudinal studies of disease, patients can experience several events across a follow-up period. Analysis of such studies can be successfully performed by multistate models. This paper considers nonparametric and semiparametric estimation of important targets in multi-state modeling, such as the transition probabilities and bivariate distribution function (for sequentially ordered events). These estimators are shown to be consistent even for data which is non-Markov. We illustrate the methods on two data sets.