(Savage Award Winner 2001/Theory and Methods)
Impartido por Luis Enrique Nieto Barajas, Departamento de EstadísticaITAM

Abstract
In a Bayesian nonparametric context, the hazard function is modelled by a stochastic process, allowing the model to have great flexibility to cope with a large range of possible behaviours of the true hazard function. We follow a Bayesian nonparametric approach and propose new prior distributions for the hazard function in univariate and bivariate cases. These priors are based on Markov processes in discrete and continuous time. Posterior distributions are obtained, and algorithms for their simulation are also shown. These models are illustrated with a full Bayesian analysis.