Doubly Regularized Cox Regression for High-Dimensional Survival Data via Cyclic Coordinate Descent

• Activity Number: 294
• Type: Topic Contributed
• Date/Time: Tuesday, August 2, 2011 : 8:30 AM to 10:20 AM
• Sponsor: IMS
• Author(s): Tongtong Wu and Sijian Wang*+
• Companies: University of Maryland and University of Wisconsin at Madison
• Keywords: Coordinate descent ; Cox regression ; High-dimensional ; Regularization ; Survival ; Variable selection

Abstract

In many scientific applications, there is a natural grouping of predictors. For example, in biological applications, assayed genes or proteins can be grouped by biological roles or biological pathways. Usually, people want to identify both important groups and important variables within selected groups. In this talk, we propose a doubly regularized Cox regression for survival data. Our regularized objective function in convex, and the method can achieve variable selection in both group level and individual level. We also developed a fast algorithm via cyclic coordinate descent method. Cyclic coordinate descent avoids matrix operations since parameters are updated one by one, which yields a very fast computing speed. It is also numerically stable due to the lack of matrix operations for large systems. We demonstrate our method and algorithm using both simulation studies and a real ovarian cancer dataset. This is a joint work with Tongtong Wu at University of Maryland.