Activity Number:
|
170
- Nonparametric Methods for Longitudinal and Survival Data
|
Type:
|
Contributed
|
Date/Time:
|
Monday, July 31, 2017 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Section on Nonparametric Statistics
|
Abstract #322875
|
|
Title:
|
Soft-Thresholding Operator for Modeling Sparse Time Varying Effects in Survival Analysis
|
Author(s):
|
Yuan Yang* and Jian Kang and Yi Li
|
Companies:
|
University of Michigan and University of Michigan and University of Michigan
|
Keywords:
|
Soft-thresholding ;
Effect sparsity ;
Survival ;
Time-varying ;
Cox
|
Abstract:
|
Precision medicine calls for the development of new methods that can identify important biomarkers and model their dynamic effects on patients' survival experiences, as existing methods are neither flexible for sparsity modeling nor computationally scalable for big data analysis. We propose a new approach to estimating sparse time-varying effects of high dimensional predictors within the Cox regression framework. As opposed to the commonly used regularization methods, we propose a new soft-thresholding operator in the space of smooth functions and use it to construct sparse and piece-wise smooth time-varying coefficients. This leads to a more interpretable model with a straightforward inference procedure. We develop an efficient algorithm for inference in the target functional space and obtain the confidence bands. We show that the proposed method enjoys good theoretical properties. The method is further illustrated and evaluated via extensive simulation studies and a data analysis of a kidney epidemiology study.
|
Authors who are presenting talks have a * after their name.