Bayesian Inference for Complex Survey Designs

• Activity Number: 433
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Lane Burgette*+ and Terrance Savitsky
• Companies: RAND Corporation and RAND Corporation
• Keywords: Bayesian analysis ; survey design ; stratification

Abstract

We propose a novel framework for fitting Bayesian models to data that were gathered according to a complex survey design. This method is generic in the sense that its implementation is essentially unaffected by the form of the outcome model, while still being fully Bayesian. Via simulation studies, we investigate the circumstances under which the method performs well, particularly in regard to small stratum sizes, and make comparisons to previous design- and model-based approaches. We also plan to investigate theoretical properties of the method, such as whether posterior consistency for a model with data gathered according to a simple random sample implies posterior consistency under this approach for handling complex survey designs.