Stable priors for Bayes Factors
Presented by Dr. Michiel DE KOCK on 8 Jul 2014 from 17:10 to 19:00
Type: Poster Presentation
Track: Track G - Theoretical and Computational Physics
Board #: G.239
Constructing reference priors for parameter spaces usually results in a proliferation of metaparameters. Bayes Factors, for example, require proper priors, and the usual class of ignorance priors then necessarily results in two metaparameters for every parameter, namely an upper and lower bound. With the goal of reducing the number of metaparameters, we replace ignorance priors by stable priors which are based on a stability property and consistent transformations between the parameters. We illustrate the procedure with an example from High-Energy Physics where we use Bayes Factors with stable priors to compare different parametrisations.
Prof. Hans Eggers firstname.lastname@example.org University of Stellenbosch