Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 414 - Risk Modeling and Regression Techniques
Type: Contributed
Date/Time: Thursday, August 12, 2021 : 2:00 PM to 3:50 PM
Sponsor: Biometrics Section
Abstract #319127
Title: Censored Broken Adaptive Ridge Regression in High-Dimension
Author(s): Jeongjin Lee* and Taehwa Choi and Sangbum Choi
Companies: Korea University and Department of Statistics, Korea University and Department of Statistics, Korea University
Keywords: Broken adaptive ridge regression; Buckley-James estimator; Accelerated failure time model; Variable selection; Coordinate descent; Survival analysis
Abstract:

Broken adaptive ridge (BAR) regression is a computationally scalable surrogate to L_0 penalized regression that performs iteratively reweighted L_2 penalized regression. It is known that the BAR regression has many appealing features, including that it compromises both L_0 and L_2 penalized regression for high-dimensional data and has the oracle properties. In this paper, we investigate the BAR procedure for variable selection in a semiparametric accelerated failure time model with complex censored data. Coupled with the Buckley-James method, we can perform the BAR-based variable selection procedure when event times are censored in a complicated manner, such as right-censoring or double-censoring. We devise a novel two-stage cyclic coordinate descent algorithm that minimizes the objective function along the direction of coordinates by the iterative estimation of the pseudo survival response and regression coefficients. Numerical studies are conducted to investigate the finite-sample performance of the proposed algorithm and an application to real data is provided as an illustration.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2021 program