Bayesian Inference in Partially Identified Models

• Activity Number: 100
• Type: Invited
• Date/Time: Monday, August 1, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Paul Gustafson*+
• Companies: University of British Columbia
• Keywords: Bayesian inference ; causal inference ; identification

Abstract

Identification can be a major issue in causal modeling contexts, and in contexts where observational studies have various limitations. Sometimes partial identification arises. In Bayesian terms, this implies that as the sample size grows, the support of the posterior distribution on the target parameter converges to a set which is smaller than the support of the prior distribution but larger than a single point. We discuss properties of Bayesian inference in partially identified models, with examples drawn from causal modeling contexts. Special attention is paid to the performance of posterior credible intervals arising from partially identified models.