Cox比例风险模型的弹性SCAD变量选择方法及应用  

Elastic SCAD Variable Selection and Its Application in Cox Model

作　　者：孙志华 谢向东 焦东方 SUN Zhi-Hua;XIE Xiang-Dong;JIAO Dong-Fang(School of Mathematical Science,Ocean University of China, Qingdao 266100, China;School of Mathematics and Statistics, Northeast Normal University, Changchun 130024, China)

机构地区：[1]中国海洋大学数学科学学院,山东青岛266100 [2]东北师范大学数学与统计学院,吉林长春130024

出　　处：《中国海洋大学学报（自然科学版）》2020年第11期129-138,共10页Periodical of Ocean University of China

基　　金：山东省自然科学基金项目(ZR2016DQ09)资助。

摘　　要：本文对基于Cox比例风险模型的弹性SCAD变量的选择方法进行了研究,证明了弹性SCAD方法在进行变量选择时具有Oracle性质、渐近正态性及组效应。通过数值模拟,将弹性SCAD与几种常见变量选择方法作了比较,得出几种不同参数设定下的变量选择结果。通过肺癌实际数据比较了弹性SCAD方法和其他常见变量选择方法。研究结果表明,弹性SCAD方法选出的变量与临床结论一致并发现了需要进一步深入研究的变量。In this paper,we studied the application of the elastic SCAD variable selection method based on the Cox proportional hazards model.Ourelastic SCAD method has the grouping effect,asymptotic consistency and the Oracle property.Wedemonstrate the performance of our method via numerical simulations.We show that the elastic SCAD method is better than other variable selections,likeLasso,Adaptive Lasso,SCAD,Elastic Net and Adaptive Elastic Net.Moreover,we apply our methodto the lung cancer data.It turns out that the elastic SCAD method performs better than other methods.Elastic SCAD is more reasonable and closer to the practical situation.

关 键 词：变量选择 弹性SCAD COX模型 组效应 Oracle性质 

分 类 号：O212.1[理学—概率论与数理统计]