|
Activity Number:
|
60
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Sunday, July 29, 2007 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #310414 |
|
Title:
|
A Flexible Variable Selection Algorithm for the Cox Model with High-Dimensional Data
|
|
Author(s):
|
Alexander Pearson*+ and Derick R. Peterson
|
|
Companies:
|
University of Rochester and University of Rochester
|
|
Address:
|
Department of Biostatistics, Rochester, NY, 14642,
|
|
Keywords:
|
Model selection ; Cox proportional hazards ; high-dimensional data
|
|
Abstract:
|
We consider the problem of variable selection with high-dimensional data in the Cox proportional hazards model framework. We propose a search algorithm that lies between forward selection and all subsets. This method uses an evolving subgroup paradigm to intelligently select variables over a larger model space than forward selection. We show that our method can yield significant improvements in the number of true variables selected and the prediction error compared with forward selection, as demonstrated with a simulations study.
|
