Blockwise Sparse Cox Regression

• Activity Number: 182
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Insuk Sohn*+ and Jinseog Kim and Sin-Ho Jung and Changyi Park
• Companies: Samsung Medical Center and Dongguk University and Duke University and University of Seoul
• Keywords: Microarray ; Clinical covariates ; Penalized cox model ; Group lasso

Abstract

The identification of influential genes (or groups of genes) and clinical covariates on the survival of patients is crucial because it can lead us to better understanding of underlying mechanism of diseases and better prediction models. However, most available statistical methods do not taken into account biological structure of gene expression and can not deal properly with categorical variables such as age, gender, and family history. In this paper, we introduce a selection method, called the blockwise sparse Cox regression (BSCR), which is an extension of the blockwise sparse regression to the Cox model. Building models based on low dimensional clinical covariates and high dimensional genomic covariates can improve the prediction. The BSCR can combine clinical and genomic covariates effectively.