|
Activity Number:
|
147
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, July 30, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #309920 |
|
Title:
|
Aspects of Bayesian Lasso Regression
|
|
Author(s):
|
Christopher Hans*+ and Steven N. MacEachern
|
|
Companies:
|
The Ohio State University and The Ohio State University
|
|
Address:
|
1958 Neil Avenue, Columbus, OH, 43210-1247,
|
|
Keywords:
|
Bayesian regression ; Gibbs sampler ; lasso ; MCMC ; prediction
|
|
Abstract:
|
Connections between Lasso estimates for linear regression and Bayesian methods usually focus on the posterior mode even though the Bayes estimate under squared error loss -- perhaps the most commonly used loss function for Bayesian estimation -- is the posterior mean. We present aspects of Bayesian Lasso regression that focus on parameter estimation via the posterior mean and on prediction of future cases via the posterior predictive distribution. Estimation of the full posterior distribution is accomplished through component-wise Gibbs sampling, including transform methods to improve mixing. While inference based on the posterior mean does not perform "variable selection" in the same way as the usual Lasso estimate, we address the question of Bayesian model comparison and selection. A link to software with efficient implementation of the methods with a simple R interface is provided.
|
