An index sampling algorithm for the bayesian analysis of a class of model selection problems

Abstract

In this paper, we present an application of the Gibbs sampler yielding an algorithm which facilitates the Bayesian analysis of a class of data partitioning or model selection problems. The resulting index sampling algorithm is applied to a univariate mean-shift outlier model for the purpose of illustration.

Keywords:

• Bayesian inference
• Gibbs sampler
• index random variables
• posterior model probabilities
• spurious and outlying observations