A Computationally Efficient Method for the Analysis of Big Survival Data

• Activity Number: 473
• Type: Invited
• Date/Time: Wednesday, August 12, 2015 : 8:30 AM to 10:20 AM
• Sponsor: WNAR
• Author(s): Kevin He and Yi Li* and Yanming Li and Ji Zhu
• Companies: University of Michigan and University of Michigan and University of Michigan
• Keywords: Big data ; Spline-based methods ; Survival analysis ; Time-varying effects

Abstract

Large-scale biomedical data that are subject to censoring frequently emerge in the era of Big Data. Time-varying effects models have become a powerful tool for describing the dynamic changes of covariate effects with long follow-up. However, the computational burden increases rapidly as the number of subjects grows, which challenges the existing statistical methods. We propose a novel application of the Quasi-Newton method with an inexact line search procedure. We show that the algorithm converges super-linearly and is, therefore, computationally efficient. The proposed method is applicable to large-scale data for which the application of existing methods is impractical. We apply the method to analyze a national kidney dataset.