|
Activity Number:
|
445
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 6, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #301922 |
|
Title:
|
Shrinkage and Model Selection with Correlated Variables via Weighted Fusion
|
|
Author(s):
|
Zhongyin J. Daye*+ and Xinge J. Jeng
|
|
Companies:
|
Purdue University and Purdue University
|
|
Address:
|
250 N. University Street, West Lafayette, IN, 47907-2066,
|
|
Keywords:
|
Elastic net ; Lasso ; Multicollinearity ; p>>n problem ; Regression ; Variable selection
|
|
Abstract:
|
Regression with correlated variables presents a challenging problem in high dimensionality. In this talk, we propose the weighted fusion, a new penalized regression and variable selection method for data with correlated variables. The weighted fusion can potentially incorporate information redundancy among related variables. When the number of predictors p is larger than the number of observations n, weighted fusion also allows the selection of more than n variables in a motivated way. We present grouping effect and consistency results for weighted fusion. Further, we demonstrate real data and simulation examples to show that weighted fusion can improve variable selection and prediction accuracy.
|
