A Review of Bayesian Model Averaging

摘要

Bayesian Model averaging is a weighted averaging method based on posterior distribution. It considers comprehensively the prior and sample information of model and parameter reduces the model uncertainty. Bayesian Model Averaging improves statistical inference accuracy and provides improved out-of-sample predictive performance. In this paper, we outlines the development of Bayesian model averaging and present research status, details of the Bayesian model averaging principle, the specific algorithm, and compares the Bayesian method with the traditional P -values.