|
Activity Number:
|
202
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #304243 |
|
Title:
|
Prior Information Allocation for Collinear Regression
|
|
Author(s):
|
Ehsan S. Soofi*+ and Nader Ebrahimi
|
|
Companies:
|
University of Wisconsin-Milwaukee and Northern Illinois University
|
|
Address:
|
P.O. BOX 742, Milwaukee, WI, 53201,
|
|
Keywords:
|
Entropy ; Information measure ; normal prior
|
|
Abstract:
|
We propose a procedure that produces optimal covariance matrix for multivariate normal prior distribution of the regression parameters such that it maximizes the sample information. The procedure consists of an optimal scheme for allocation of the total prior variance to the principal component transforming of the model parameters. The optimal solution is in terms of the eigenvalues, which encapsulate the collinearity effects on estimation, and a tuning parameter, which is interpreted as the prior to model precision in the linear regression. This procedure is applicable to Bayesian analysis of linear models with multivariate normal prior and posterior distributions, and other important models such as logit and proportional hazard models where the posterior distribution is approximately normal.
|
