Statistics and Its Interface

Volume 10 (2017)

Number 3

Efficient feature screening for ultrahigh-dimensional varying coefficient models

Pages: 407 – 412

DOI: https://dx.doi.org/10.4310/SII.2017.v10.n3.a5

Authors

Xin Chen (Department of Statistics and Applied Probability, National University of Singapore)

Xuejun Ma (College of Applied Sciences, Beijing University of Technology, Beijing, China)

Xueqin Wang (Southern China Center for Statistical Science, School of Mathematics, Zhongshan School of Medicine, Xinhua College, Sun Yat-Sen University, Guangzhou, Guangdong Province, China)

Jingxiao Zhang (Center for Applied Statistics, School of Statistics, Renmin University of China, Beijing, China)

Abstract

Feature screening in ultrahigh-dimensional varying coefficient models is a crucial statistical problem in economics, genomics, etc. Current methods not only suffer from circumstances when the models involve multiple index variables or group predictor variables, but also cannot handle nonlinear varying coefficient models. To address these reallife scenarios efficiently, we develop a screening procedure for ultrahigh-dimensional varying coefficient models utilizing conditional distance covariance (CDC). Extensive simulation studies and two real economic data examples show the effectiveness and the flexibility of our proposed method.

Keywords

ultrahigh-dimensionality, varying coefficient models, multiple index variables, group variables, conditional distance covariance

2010 Mathematics Subject Classification

62G08, 62G20, 62H20

Full Text (PDF format)

Published 31 January 2017