Group-wise semiparametric modeling: A SCSE approach

JOURNAL OF MULTIVARIATE ANALYSIS

卷:152

Article

Song, Song¹  Zhu, Lixing^2,3 

[1] Univ Texas Austin, Austin, TX 78712 USA

[2] Suzhou Univ Sci & Technol, Suzhou, Jiangsu, Peoples R China

[3] Hong Kong Baptist Univ, Hong Kong, Hong Kong, Peoples R China

关键词: Covariance estimation;    Regularization;    Sparsity;    Thresholding;    Semiparametrics;    Variable clustering;   

DOI  :  10.1016/j.jmva.2016.07.006

来源: Elsevier

【 摘 要 】

This paper is motivated by the modeling of a high-dimensional dataset via group-wise information on explanatory variables. A three-step algorithm is suggested for group-wise semiparametric modeling: (i) screening to reduce dimensionality; (ii) clustering according to grouped explanatory variables: (iii) sign-constraints-based estimation for coefficients to produce meaningful interpretations. As a justification, under the setup of m-dependent and beta-mixing processes, the interplay between the estimator's convergence rate and the temporal dependence level is quantified and a cross-validation result about the resampling scheme for threshold selection is also proved. This method is evaluated in finite-sample cases through a Monte Carlo experiment, and illustrated with an analysis of the US consumer price index. (C) 2016 Elsevier Inc. All rights reserved.

【 授权许可】

Free