Gateau Differential-Based Boosting for Time-Varying Survival Models

• Activity Number: 538
• Type: Invited
• Date/Time: Wednesday, August 6, 2014 : 2:00 PM to 3:50 PM
• Sponsor: International Indian Statistical Association
• Author(s): Yi Li*+ and Kevin He and Ji Zhu
• Companies: University of Michigan and University of Michigan and University of Michigan
• Keywords: Boosting ; High dimensional data ; Survival models ; Time-varying effects ; Variable selection

Abstract

Survival models with time-varying effects provide a flexible framework for modeling the effects of covariates on event times. In view of existing work that is often focused on time-varying models with relatively low dimensionality, we propose a new Gateau differential-based boosting procedure for simultaneously selecting and automatically determining the functional form of covariates. Specifically, our procedure allows that in each boosting learning step only the best-fitting base-learner (and therefore the most informative covariate) is added to the predictor, and consequently encourages sparsity. In addition, our method controls smoothness, which is crucial for improving the predictive performance. The performance of the proposed method is examined by simulations and by applications to analyze the multiple myeloma data and national kidney transplant data.