|
Activity Number:
|
449
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 6, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #301176 |
|
Title:
|
A General Framework for Bilevel Variable Selection
|
|
Author(s):
|
Patrick Breheny*+ and Jian Huang+
|
|
Companies:
|
The University of Iowa and The University of Iowa
|
|
Address:
|
, , , , , ,
|
|
Keywords:
|
Lasso ; Penalized Regression ; Group lasso ; SCAD ; Variable Selection ; Group bridge
|
|
Abstract:
|
Penalized methods are very useful in sparse regression problems. In many applications, variables have a grouping structure that can be incorporated into the penalty. Here, we introduce a general framework that allows penalties such as ridge, lasso, and SCAD to be combined independently at the group and individual levels, and present algorithms to fit these models. This framework incorporates two existing approaches, group bridge and group lasso. The proposed methods are capable of simultaneous selection at the group and individual variable levels and can be applied to linear and general regression models. Simulations indicate that these group approaches are superior to penalties that ignore grouping structures and provide insight into choosing penalties that best reflect study goals. Finally, we apply these approaches to a genetic association study of age-related macular degeneration.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2008 program |
