Online Program Home
  My Program

Abstract Details

Activity Number: 470 - Bayes Theory and Foundations
Type: Contributed
Date/Time: Wednesday, August 2, 2017 : 8:30 AM to 10:20 AM
Sponsor: Section on Bayesian Statistical Science
Abstract #324632
Title: Robust Subjective Bayesian Hypothesis Testing by Calibrating Prior Probabilities
Author(s): Dan Spitzner*
Companies: University of Virginia
Keywords: Bayes factors ; hypothesis testing ; default priors ; Bayesian information criterion ; multiple testing ; variable selection
Abstract:

In testing problems, calibrated prior probabilities may be used to modify the framework for Bayes factors in order to reduce sensitivity to scale parameters of the prior, and are thus suitable for use with vague priors. The framework has a particular advantage for subjective Bayesian analysis in that both the continuous and discrete portions of the prior are easily interpreted, which simplifies the elicitation of prior probabilities. The challenge of using calibrated prior probabilities is ambiguity in the calibration rule. This presentation proposes to use existing default priors in a novel way, as anchors the calibration rather than a prior specification in itself. The reasoning and interesting consequences of this approach is discussed, as well as its advantages to interpretation and such practical matters as multiple testing.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2017 program

 
 
Copyright © American Statistical Association