Bayesian analysis of a mixture of two components of different families of distributions under different censoring schemes

Abstract

Finite mixtures models have received considerable attention in areas of survival analysis and reliability in recent years, from analysis in both the methodological development and multifarious applications. This thesis discuss the statistical inference of heterogeneous population model by using two component mixture model when the data are of different censoring schemes. We consider the maximum likelihood estimation and Bayes estimation of parameters assuming informative and non-informative priors under symmetric and asymmetric loss functions. In some cases three different approximation methods are used for Bayesian computation, importance sampling method, Lindley approximation and Tierney and Kadane approximation. We perform Monte Carlo simulation to compare the performance of the different methods. The Bayes prediction intervals are also determined.