Forward Stagewise Shrinkage and Addition for High and Ultra-High Dimensional Censored Regression

• Activity Number: 299
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 8:30 AM to 10:20 PM
• Sponsor: Biometrics Section
• Author(s): Zifang Guo*+ and Wenbin Lu and Lexin Li
• Companies: North Carolina State University and North Carolina State University and North Carolina State University
• Keywords: Adaptive LASSO ; boosting ; forward stagewise regression ; proportional hazards model ; variable selection

Abstract

Despite the thriving development of variable selection methods in recent years, modeling and selection of high and ultrahigh dimensional censored regression remain challenging. When the number of predictors p far exceeds the number of observational units n, computations of many methods become difficult or even infeasible. Censoring of the outcome variable adds further complications. In this article, we propose a forward stagewise shrinkage and addition method for simultaneous model estimation and variable selection in Cox proportional hazards models with high and ultrahigh dimensional covariates. Our proposal extends a popular statistical learning technique, the boosting method, by explicitly performing variable selection and substantially reducing the number of iterations for algorithm convergence. It also inherits the flexible nature of the boosting and is straightforward to extend to nonlinear Cox models. Our intensive numerical analyses demonstrate that the new method enjoys an equally competitive performance as the best players of the existing solutions in Cox models with p < n, whereas it achieves a considerably superior performance than the alternative solutions when p > n.