The use of random-effect models for high-dimensional variable selection problems

被引：11

作者：

Kwon, Sunghoon ^{[1
]}

Oh, Seungyoung ^{[2
]}

Lee, Youngjo ^{[2
]}

机构：

[1] Konkuk Univ, Dept Appl Stat, Seoul, South Korea

[2] Seoul Natl Univ, Dept Stat, Seoul, South Korea

来源：

COMPUTATIONAL STATISTICS & DATA ANALYSIS | 2016年 / 103卷

基金：

新加坡国家研究基金会;

关键词：

Generalized linear model; Hierarchical likelihood; High-dimension; Random effect; Unbounded penalty; Variable selection; NONCONCAVE PENALIZED LIKELIHOOD; DIVERGING NUMBER; LINEAR-MODELS; REGRESSION; REGULARIZATION; CLASSIFICATION; MICROARRAYS; SHRINKAGE; LASSO;

D O I：

10.1016/j.csda.2016.05.016

中图分类号：

TP39 [计算机的应用];

学科分类号：

081203 ; 0835 ;

摘要：

We study the use of random-effect models for variable selection in high-dimensional generalized linear models where the number of covariates exceeds the sample size. Certain distributional assumptions on the random effects produce a penalty that is non-convex and unbounded at the origin. We introduce a unified algorithm that can be applied to various statistical models including generalized linear models. Simulation studies and data analysis are provided. (C) 2016 Elsevier B.V. All rights reserved.

引用

页码：401 / 412

页数：12

共 50 条

[1]

[Anonymous], 2006, Journal of the Royal Statistical Society, Series B

[2]

[Anonymous], P NIPS 2012 OPT WORK

[3]

Bühlmann P, 2008, ANN STAT, V36, P1534, DOI 10.1214/07-AOS0316A

[4] AN INTRODUCTION TO EMPIRICAL BAYES DATA-ANALYSIS [J].