Variable Selection in Competing Risks Using the L1 Penalized Cox Model

• Activity Number: 382
• Type: Contributed
• Date/Time: Tuesday, August 4, 2009 : 2:00 PM to 3:50 PM
• Sponsor: Section on Risk Analysis
• Author(s): Xiangrong Kong*+ and Kellie J. Archer
• Companies: Johns Hopkins University and Virginia Commonwealth University
• Keywords: Competing risks ; Variable selection ; Penalized model ; Cox model

Abstract

One situation in survival analysis is the failure of an individual can happen due to one of multiple causes. Survival data generated in this scenario are referred to as competing risks data. There may be predictor variables associated with the risks raised from the different causes being studied. The same variable can have different degrees of influence on the different risks. Given a set of predictors, it is of interest to develop a statistical method to identify the subset of variables that are significantly associated with the risk corresponding to each cause. One important feature of our method is the ability to perform variable selection in situations where the number of predictor variables is larger than the sample size. The method was applied on a real data set from an NIH funded project to identify genes that may be relevant to tumor progression in HCV infected patients with HCC.