High-dimensional rank tests for sphericity

JOURNAL OF MULTIVARIATE ANALYSIS

卷:155

Article

Feng, Long¹  Liu, Binghui¹ 

[1] Northeast Normal Univ, Sch Math & Stat, Key Lab Appl Stat MOE, Changchun 130024, Peoples R China

关键词: Asymptotic normality;    High dimension;    Rank test;    Spatial rank;    Sphericity test;   

DOI  :  10.1016/j.jmva.2017.01.003

来源: Elsevier

【 摘 要 】

In recent years, procedures for testing distributional sphericity have attracted increased attention, especially in high-dimensional settings. A prominent problem in this context is the development of robust and efficient test statistics. In this paper, we propose two novel rank tests inspired by Spearman's rho and Kendall's tau for high-dimensional problems. Due to the blessing of dimension, estimation of masses of nuisance parameters is avoided, which allows our procedures to work in arbitrary large dimension. The asymptotic normality of the proposed tests is established for elliptical distributions and their performance is investigated over a wide range of simulation set-ups. (C) 2017 Elsevier Inc. All rights reserved.

【 授权许可】

Free