Abstract #301558


The views expressed here are those of the individual authors
and not necessarily those of the ASA or its board, officers, or staff.


Back to main JSM 2002 Program page



JSM 2002 Abstract #301558
Activity Number: 5
Type: Invited
Date/Time: Sunday, August 11, 2002 : 2:00 PM to 3:50 PM
Sponsor: Section on Bayesian Stat. Sciences*
Abstract - #301558
Title: Dilution Priors for Model Uncertainty
Author(s): Edward George*+
Affiliation(s): University of Pennsylvania
Address: 3620 Locust Walk, Philadelphia, Pennsylvania, 19104, USA
Keywords: model space priors ; objective Bayesian analysis ; Voronoi tesselation ; model averaging ; model selection
Abstract:

In model uncertainty setups, some models may be very similar to others. For example, this would occur with the uncertainty about which of several highly correlated variables to include in a linear regression. In such setups, it is tempting to use characterize ignorance by a uniform prior which assigns equal probability to each model under consideration. Unfortunately, such a prior may not be uniform on "neighborhoods of models," especially when there are regions of high redundancy in the model space. By skewing the posterior, such a prior can denigrate the predictive potential of model averaging. Dilution priors rectify this problem by assigning probability more uniformly to model neighborhoods. Such priors do not require any subjective inputs. This talk will focus on the construction of dilution priors for two setups: Bayesian variable selection for the linear model and Bayesian CART model selection.


  • The address information is for the authors that have a + after their name.
  • Authors who are presenting talks have a * after their name.

Back to the full JSM 2002 program

JSM 2002

For information, contact meetings@amstat.org or phone (703) 684-1221.

If you have questions about the Continuing Education program, please contact the Education Department.

Revised March 2002