Duke University

presents:

Bruno Sansó

Duke University

and

University of Simón Bolívar

"Intrinsic Bayes Factor and Intrinsic Priors: Some Robustness Issues"

Abstract:

Bayesian model comparisons are known to be non-robust with respect to large classes of proper priors and are undetermined when improper priors are employed.

The Intrinsic Bayes factor (IBF) is a general automatic procedure for model comparison proposed in Berger and Pericchi (1993) which overcomes the difficulties that arise when improper priors are employed. An appealing justification of the IBF is that it corresponds to the actual Bayes factors of particular priors; such priors can be obtained as the solution of two functional equations generated from the asymptotic properties of the IBF and are called the instrinsic priors. In this talk I will briefly review the main ideas of IBFs and intrinsic priors and then I will consider issues related to the robustness of the IBF for some basic inferential problems, in the nested model situation. I will consider robustness with respect to the class of intrinsic priors and robustness with respect to the choice of the non-informative priors for each model.

Friday, November 17, 1995

11:45 - 12:45

116 Old Chemistry Building