Intrinsic Analysis of Gaussian and Latent Gaussian Data

• Activity Number: 676
• Type: Topic Contributed
• Date/Time: Thursday, August 8, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Andrew Womack*+
• Companies: University of Florida
• Keywords: Intrinsic Prior ; Model Selection ; Bayes Factors ; RJMCMC ; Linear regression ; Probit regression

Abstract

In this talk, we present recent results on the analysis of Gaussian and probit regression models using the intrinsic methodology. This includes direct sampling techniques for posterior distributions and the construction of point and interval estimates. We also discuss extensions of the methodology including ordinal probit models and structured covariate selection. For these extensions, we present a RJMCMC algorithm for sampling from the model averaged posterior distribution as well as Rao-Blackwell estimators for posterior model probabilities. We demonstrate the RJMCMC algorithm in an application to structured GWAS using gene ontology data.