|
Title
|
Room
|
|
! Flexible Bayesian Regression Analysis
|
H-State/Club
|
|
Date / Time
|
Sponsor
|
Type
|
|
08/06/2001
10:30 AM
-
12:20 PM
|
SSC, ENAR
|
Invited
|
|
Organizer:
|
Bertrand Clarke, University of British Columbia
|
|
Chair:
|
Bertrand Clarke, University of British Columbia
|
|
Discussant:
|
|
|
Floor Discussion
|
12:05 PM
|
Description
In a large-data era, fitting flexible (nonlinear and nonadditive) models to datasets with many predictors is an increasingly common statistical task. Indeed, some of the research activity in this area has been subsumed by the expanding fields of machine learning and data mining. A Bayes approach to such fitting can have several appealing features: explicit penalization of complexity via a prior, the power of MCMC algorithms, a full accounting of uncertainty about the model structure, and the ability to average models for improved predictive performance. This session would highlight various approaches (quite different, though all Bayesian) to this problem.
|
|
