|
Activity Number:
|
318
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Technometrics
|
| Abstract - #304888 |
|
Title:
|
Simultaneous Variable Selection
|
|
Author(s):
|
Berwin A. Turlach*+ and William N. Venables and Stephen J. Wright
|
|
Companies:
|
The University of Western Australia and CSIRO Mathematical and Information Sciences and University of Wisconsin-Madison
|
|
Address:
|
School of Mathematics and Statistics (M019), Crawley, WA, 6009, Australia
|
|
Keywords:
|
subset selection ; multivariate regression ; variable selection ; interior point algorithm ; solution path algorithm
|
|
Abstract:
|
We propose a new method for selecting a common subset of explanatory variables where the aim is to model several response variables. The idea is a natural extension of the LASSO technique and based on minimizing the (joint) residual sum of squares while constraining the parameter estimates to lie within a suitable polyhedral region. We briefly comment on other constraints that may be imposed to achieve simultaneous variable selection. The properties of the convex programming problem resulting from our approach are analyzed for the special case of an orthonormal design. For the general case, we develop an efficient interior point algorithm and also describe an algorithm that calculates the complete solution path. Time permitting, the method will be illustrated on various datasets.
|
