Fractional Bayes Factors for Model Comparison

摘要: Bayesian comparison of models is achieved simply by calculation posterior probabilities the themselves. However, there are difficulties with this approach when prior information about parameters various weak. Partial Bayes factors offer a resoIution problem setting aside part data as training sampIe. The sampIe used to obtain an initiaI informative distribution in each model. Model then based on factor calculated from remaining data. Properties partial discussed, particularly context weak information, and they found have advantages over other proposed methods model comparison. A new variant factor, fractional advocated grounds consistency, simplicity, robustness coherence