Spring Speaker 2008

The Purdue Statistics Graduate Student Organization (GSO) is pleased to have Professor James O. Berger as this year's spring speaker. Professor Berger will speak on Thursday, February 21, 2008 at 4:30pm in MATH 175. The title of his talk will be "A Review of Surprises Encountered in Bayesian Model Selection." The talk is joint with the Department of Statistics Research Colloquium. Refreshments will be served in HAAS 101 at 4:15pm.

Professor James O. Berger is The Arts and Sciences Professor of Statistics at Duke University and Director of the Statistical and Applied Mathematical Sciences Institute (SAMSI). Before moving to Duke, he was a faculty member in the Purdue University Department of Statistics from 1974-1997. Well known for his work in Bayesian statistics, Professor Berger's research interests also include the foundations of statistics, statistical decision theory, model selection, and simulation.

To read more about Professor Berger, please visit his website.

Thursday, February 21, 2008
4:30PM in MATH 175

Professor James O. Berger

The Arts and Sciences Professor of Statistics, Duke University
and
Director of the Statistical and Applied Mathematical Sciences Institute (SAMSI)

Joint with Department of Statistics Research Colloquium

A Review of Surprises Encountered in Bayesian Model Selection

Abstract
This talk reviews the following ideas, all of which I at one time thought to be true, but now think false.

• Use of p-values is better than fixed alpha-level testing, since p-values are conditional on the data.
• Frequentist testing and Bayesian testing are incompatible; for instance, Bayes tests do not depend on the stopping rule in sequential settings while frequentist tests do so depend, necessitating "spending alpha" for looks at the data.
• The best single model to a Bayesian is the highest posterior probability model.
• Model selection priors cannot be derived from the data.
• Only a relatively small number of models will typically receive significant posterior probability (or other "weight"), and hence description of model uncertainty can focus on a few best models.

Again, I now view all these statements to be false, and will discuss why. Many of these issues will be illustrated through an example involving high-energy physics.