Reference and Matching Priors in Objective Bayesian Analysis

Abstract

Because of the needs in many applied Bayesian analyses, objective Bayesian method attracts more attention lately among open-minded Bayesians and frequentists. In this talk, we will focus on two issues: (1) Bayesian reference prior analysis; and (2) posterior matching prior analysis. The algorithms of calculating those priors will be discussed. Also, those methodologies will be applied to statistical models such as statistical calibration models under different settings. It will be shown that most of those objective Bayesian analyses result in desired frequentist properties. Furthermore, since the methods are used on multi-dimensional parameter spaces, they usually give satisfactory results in eliminating nuisance parameters.