|
Activity Number:
|
449
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 6, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #300503 |
|
Title:
|
Boosting Accelerated Failure Time Models for Survival Data with High-Dimensional Covariates
|
|
Author(s):
|
Zhu Wang*+ and C. Y. Wang
|
|
Companies:
|
Yale University and Fred Hutchinson Cancer Research Center
|
|
Address:
|
950 Campbell Avenue, West Haven, CT, 06516,
|
|
Keywords:
|
Boosting ; Accelerated failure time model ; Buckley-James estimator ; Censored survival data ; Variable selection
|
|
Abstract:
|
We consider variable selection for the accelerated failure time models with censored survival data. With high-dimensional covariates, boosting has being successfully applied to regression and classification problems for building accurate predictive models and conducting variable selection simultaneously. However, only limited effort has been made with boosting for the accelerated failure time models when censoring occurs. In this paper, boosting with componentwise linear least squares for model selection is presented for right censored time-to-event data. To accommodate censoring, we consider the mean imputation and Buckley-James method. The proposed methods are evaluated and the Buckley-James method is shown to be comparable with the weighted least squares boosting and weighted LASSO using simulated data. The methods are illustrated with microarray gene expression data DLBCL.
|
