Date de l'évènement :
Lieu de l'évènement :
Organisé par :
Two classes of statistical models for the analysis of multi-state models will be reviewed: intensity-based models and marginal models. In the former class of models, one aims at modeling the entire multi-state process and all its transition intensities while, in the latter class, one focuses on direct estimation of certain marginal features, such as state occupation probabilities. We will argue that the latter class is particularly relevant in randomized studies where the aim is a comparison of two treatment arms. Examples from medical research will illustrate the methods.
key words: Multi-state model, survival analysis, transition intensity, marginal parameter, pseudo-observation