Model-Free Prediction of Survival Probability with High-Dimensional Covariates

• Activity Number: 253
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Yuan Geng*+ and Wenbin Lu and Hao Helen Zhang
• Companies: North Carolina State University and North Carolina State University and North Carolina State University
• Keywords: Survival probability prediction ; High-dimensional covariates ; Nonparametric ; Model-free ; Weighted support vector machine ; Inverse censoring probability weight

Abstract

High dimensional data like gene expression data have been widely used as predictors to estimate survival in cancer study. The traditional semiparametric methods like proportional hazard model or propotional odds model have difficulty in handling high dimensional covariates. They also suffer the systematic bias in case of model misspecification. We propose a nonparametric weighted support vector machine (SVM) method using the inverse censoring probability weight (ICPW), which is model-free and thus robust of model specification. Furthermore, the proposed method is also reliable when the censor rate is high. The simulations and the real data analysis show that the proposed method presents a better performance in survival probability prediction for the high-dimensional data compared with the traditional methods.