|
Activity Number:
|
410
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #310189 |
|
Title:
|
Bayesian Model Selection by Smoothly Clipped Continuous Priors
|
|
Author(s):
|
Sounak Chakraborty*+
|
|
Companies:
|
University of Missouri-Columbia
|
|
Address:
|
134 K Middlebush Hall, Columbia, MO, 65201,
|
|
Keywords:
|
model selection ; SCAD ; Bayesian analysis ; classificartion ; SVM
|
|
Abstract:
|
Model selection is present in the Bayesian literature for over a long time. The stochastic search variable selection or Variable selection through BIC has been extensively used for simultaneously choosing variables when we have a huge number of explanatory variables to decide from. A very apt example is the need for a sophisticated variable selection technique in gene expression microarray experiment. Under the regression setup the LASSO has been widely used for this purpose. Which has a direct relationship with a Bayesian model with double exponential prior. In this paper we talk about a new king of a continuous priors based on SCAD or smoothly clipped penalty function for variable selection. This SCAD prior can be used under both regression and classification setup and can be directly linked with Bayesian SVM for simultaneous classification and class prediction.
|
