Bayesian model selection and prediction with empirical applications discussion

PIC (F) criterions are related to parallel developments in the Bayesian literature. Phillipsʹ concept of a ‘Bayes model’ is compared to the traditional definition of Bayesian models. It is argued that, while the PIC (F) criterion is an important ‘design’ criterion, it does not provide an alternative ‘model’ in itself. A related distinction is drawn between the concepts of ‘model choice’ and encompassing.