|
Activity Number:
|
491
|
|
Type:
|
Invited
|
|
Date/Time:
|
Thursday, August 2, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
IMS
|
| Abstract - #307709 |
|
Title:
|
Prior Elicitation and Variable Selection in Regression Models with High-Dimensional Data
|
|
Author(s):
|
Joseph G. Ibrahim*+ and Mayetri Gupta
|
|
Companies:
|
The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill
|
|
Address:
|
Department of Biostatistics, Chapel Hill, NC, 27614,
|
|
Keywords:
|
Generalized Linear Models ; Survival Analysis ; g-prior ; Microarray Data
|
|
Abstract:
|
One of the most important modern day challenges in analyzing high dimensional data (such as microarray data) jointly with continuous, discrete, or time-to-event data is that one immediately encounters the p > n problem, in which the number of covariates in the regression model greatly exceeds the number of subjects. In this talk, we develop a methodology for the specification of a class of prior distributions for generalized linear models and survival models that accommodate the p > n paradigm. The class of proper prior distributions are based on the generalization of the g-prior as well as having a 'ridge parameter' that facilitates propriety for p > n. The resulting prior has the flavor and operating characteristics of ridge regression. Various properties of the prior are discussed and a real dataset is analyzed to illustrate the proposed methodology.
|
