Modelos de sobrevivência bivariados baseados na cópula FGM : uma abordagem bayesiana

Abstract

In this work we present a Bayesian analysis for bivariate survival data in the presence of a covariate and censored observations. We propose a bivariate distribution for the bivariate survival times based on the Farlie-Gumbel-Morgenstern (FGM) copula to model data with weak dependence. Some survival models with and without cure rate have been assumed for the marginal distributions. For inferential purpose a Bayesian approach via Markov Chain Monte Carlo (MCMC) was considered. Further, some discussions on model selection criteria are given and comparisons with other copula models were performed. To detect influential observations in the data we consider a Bayesian case deletion influence diagnostics based on the -divergence. The OpenBUGS and R systems were used to simulate samples of the posterior distribution. Numerical illustrations are presented considering artificial and real data sets.