|
Activity Number:
|
96
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, July 30, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #308416 |
|
Title:
|
A Variable Selection Method for Linear Models Using Modified Zellner's Prior
|
|
Author(s):
|
Arun Krishna*+ and Sujit Ghosh and Howard D. Bondell
|
|
Companies:
|
North Carolina State University and North Carolina State University and North Carolina State University
|
|
Address:
|
4700 Westgrove Street, Raleigh, NC, 27606,
|
|
Keywords:
|
Variable Selection ; Zellner's g-priors ; Linear Model
|
|
Abstract:
|
A common method to solve variable selection problems in linear models is minimizing a penalized sum of squares, where most of the methods differ in the penalty function. Recent methods are mostly based on penalty functions including the norms of parameter estimates, or combination of them. Ridge Regression, LASSO and Elastic Net can be considered as popular examples of these shrinkage based methods. These problems can be considered under a Bayesian framework where the log of the prior density of the parameters act as penalty functions. A suitable modification of the popular Zellner's prior for regression coefficient is proposed and the penalty function is allowed to depend on the design matrix. Performances of the new prior (and hence new penalty function) are compared with some of the current variable selection methodologies using simulated and real data.
|
