Censored Rank Independence Screening for High-Dimensional Survival Data

• Activity Number: 585
• Type: Topic Contributed
• Date/Time: Wednesday, August 7, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Wenbin Lu*+ and Rui Song and Shuangge Ma
• Companies: Department of Statistics, North Carolina State University and North Carolina State University and Yale University
• Keywords: High-dimensional survival data ; Rank independence screening ; Sure screening property

Abstract

High-dimension is an important feature of modern statistical applications, where the dimen-sionality of covariates is potentially much larger than sample size. Correlation screening (Fan and Lv, 2008) was established to be effective to reduce the dimension for such data while achieve the sure screening property for the linear models, that is, all the active variables can be retained with overwhelming probability. Screening based on Pearson's correlation, however, does not perform well with contaminated covariates and censored outcomes. In this paper, we study censored rank independence screening with high dimensional survival data. The proposed method is robust to predictors with outliers. It works for a general class of survival models. Its sure screening prop-erty is established. Simulations and a real data analysis demonstrate that the proposed method has very competitive performance for survival data with moderate sample size and very high dimensional predictors, particularly with contaminated predictors.