Name: Siddhartha Chib
From: Washington University in Saint Louis
Title: Bayes from Moments
Abstract: This talk is a summary of recent work, developed in Chib, Shin and Simoni (2018,2019) on Bayesian inference when the unknown distribution of the outcomes is specified up to a set of over-restricted unconditional or conditional moments, some of which may be mis-specified. The likelihood for the analysis is the exponentially titled empirical likelihood (ETEL), which, unlike the closely related empirical likelihood, has Bayesian underpinnings. Under regularity conditions, the posterior distribution under the ETEL function is shown to satisfy the Bernstein-von-Mises theorem, even under misspecification of the moments. We also discuss the computation and performance of the marginal likelihood for comparing such moment condition models, and provide large sample model consistency results. Several examples are provided, along with the outlines of ongoing work.