CMStatistics 2016: Start Registration
View Submission - CMStatistics
B0918
Title: Bayesian inference and identification issues in a flexible promotion time cure model Authors:  Philippe Lambert - Universite de Liege / Universite catholique de Louvain (Belgium) [presenting]
Abstract: The promotion time model is a special case of cure survival models where an unidentified proportion of subjects will never experience the event of interest whatever the duration of the follow-up. We focus our interest on a flexible formulation where the baseline hazard for the susceptible subjects is specified using Bayesian $P$-splines. Identification issues arise under an insufficiently long follow-up when the same covariates enter the model specification for the probability to be cured and the time-to-event part. An original proposal to automatically identify such situations will be described and illustrated on real data. Inference will be made in a Bayesian framework using MCMC and compared with results based on Laplace approximations.