False Selection Rate Methods in the Cox Proportional Hazards Model
No Thumbnail Available
Files
Date
2006-07-27
Authors
Journal Title
Series/Report No.
Journal ISSN
Volume Title
Publisher
Abstract
Variable selection methods are useful for distinguishing informative variables from uninformative variables. Many variable selection methods have been studied in linear regression models. Some methods have been extended to the Cox proportional hazards model based on the partial likelihood function. The False Selection Rate (FSR) variable selection method introduced by Wu, Boos and Stefanski (2006) is a new method applied to forward selection that controls the false selection rate of uninformative variables in the model by adding a number of phony variables to the original variables and calculating the proportions of the phony variables selected. In this work, we adapt the False Selection Rate procedure with the forward selection method (Forward-FSR) to the Cox proportional hazards model. In addition, we propose a new approach to estimate the tuning parameters in the penalty functions of the LASSO and the SCAD methods based on the False Selection Rate variable selection criterion. Under the Cox proportional hazards model, the three FSR methods, Forward-FSR, LASSO-FSR, and SCAD-FSR, are compared to forward selection with information criteria, to the LASSO and to the SCAD methods in simulation studies.
Description
Keywords
SCAD-FSR, model selection, Forward-FSR, false selection rate (FSR), LASSO-FSR, the Cox model
Citation
Degree
PhD
Discipline
Statistics
