Dynamic Bayesian Models for Survival Data

Abstract

Dynamic models are proposed for the study of survival data with explanatory variables whose effects change through time. The parameters modelling these effects are allowed to vary between time intervals and a system equation provides the stochastic link for adjacent values. Sequential analysis is used, based on a factorization of the likelihood over the time intervals. The updating equations are obtained via the dynamic generalized modelling approach of West, Harrison and Migon. Predictive features for follow‐up studies and analysis of new observations are obtained and some numerical applications are provided.