Empirical Bayes Model Averaging Under Model Misfit

• Activity Number: 263
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Junyan Wang* and Chris Hans and Mario Peruggia
• Companies: The Ohio State University and The Ohio State University and The Ohio State University
• Keywords: BMA ; Zellner's g prior ; local shrinkage ; out-of-sample prediction error ; model misspecification ; regression modeling

Abstract

A routine application of Bayesian Model Averaging (BMA) assumes implicitly that the data satisfy the regression modeling assumptions. We study the performance of BMA with respect to out-of-sample prediction accuracy when these assumptions are violated. Zellner's g prior remains popular in Bayesian regression modeling, and treatment of the parameters in the prior are directly connected to the quality of posterior inference and prediction. We investigate methods of empirical estimation of tuning parameters in the prior that attempt to attenuate the impact of model misfit on prediction accuracy. Our initial work has focused on optimal prediction under the g prior in the presence of influential cases. The central theme of the work is the development of prescriptive guidelines for optimal choices of g under various violations of the model assumptions.