Use of Mixture of Distributions for the Error in Regression Models for Lifetime Data

摘要

In this paper we present a Bayesian analysis of location-scale regression modelsassuming standard lifetime distributions and an additional error term with amixture of normal distributions. Considering a censored lifetime data set, we useGibbs sampling with Metropolis-Hastings algorithm to get Bayesian quantitiesof interest. The proposed regression mo del gives a great .exibility to fit lifetimedata as we see in an example