Relatore: prof. Ioannis Ntzoufras 

Athens University of Economics and Business (AUEB), Department of Statistics

Abstract

In this talk I will provide  a review of prior distributions for objective Bayesian model comparisons. The general principles, criteria and tools/mechanisms that can be used in order to ensure a sensible Bayesian model comparison/selection procedure under the absence of any prior information will be presented and discussed. Focus will be given in the most popular model selection case: the variable selection problem.  Some recent contributions in the area of objective priors on model space will be also discussed. Hopefully, after this talk you will have some guidance how to work to build sensible priors for Bayesian model comparisons.