Assigning a value to a power likelihood in a general Bayesian model

Abstract

Bayesian robustness under model misspecification is a current area of active research. Among recent ideas is that of raising the likelihood function to a power. In this paper we discuss the choice of appropriate power and provide examples.