Hypothesis Testing for the Covariance Matrix in High-Dimensional Transposable Data with Kronecker Product Dependence Structure

Anestis Touloumis, John C. Marioni, Simon Tavaré

Research output: Contribution to journalArticle

Abstract

The matrix-variate normal distribution is a popular model for high-dimensional transposable data because it decomposes the dependence structure of the random matrix into the Kronecker product of two covariance matrices, one for each of the row and column variables. However, there is a lack of hypothesis testing procedures for the row or column covariance matrix in high-dimensional settings. Tests for assessing the sphericity, identity and diagonality hypothesis for the row (column) covariance matrix in high-dimensional settings while treating the column (row) dependence structure as a ‘nuisance’ parameter are introduced. The proposed tests are robust to normality departures provided that the Kronecker product dependence structure holds. In simulations, the proposed tests appear to maintain the nominal level and they tended to be powerful against the alternative hypotheses tested. The utility of the proposed tests is demonstrated by analyzing a microarray and an electroencephalogram study. The proposed testing methodology has been implemented in the R package HDTD.
Original languageEnglish
JournalStatistica Sinica
DOIs
Publication statusAccepted/In press - 28 Oct 2019

Fingerprint

Kronecker Product
Dependence Structure
High-dimensional Data
Hypothesis Testing
Covariance matrix
High-dimensional
Sphericity
Nuisance Parameter
Random Matrices
Microarray
Normality
Categorical or nominal
Gaussian distribution
Decompose
Testing
Methodology
Alternatives
Simulation
Model

Keywords

  • Covariance matrix
  • high-dimensional settings
  • hypothesis testing
  • matrix-valued random variables
  • transposable data

Cite this

@article{9a6885959aa54ec4ae935562cd57d78b,
title = "Hypothesis Testing for the Covariance Matrix in High-Dimensional Transposable Data with Kronecker Product Dependence Structure",
abstract = "The matrix-variate normal distribution is a popular model for high-dimensional transposable data because it decomposes the dependence structure of the random matrix into the Kronecker product of two covariance matrices, one for each of the row and column variables. However, there is a lack of hypothesis testing procedures for the row or column covariance matrix in high-dimensional settings. Tests for assessing the sphericity, identity and diagonality hypothesis for the row (column) covariance matrix in high-dimensional settings while treating the column (row) dependence structure as a ‘nuisance’ parameter are introduced. The proposed tests are robust to normality departures provided that the Kronecker product dependence structure holds. In simulations, the proposed tests appear to maintain the nominal level and they tended to be powerful against the alternative hypotheses tested. The utility of the proposed tests is demonstrated by analyzing a microarray and an electroencephalogram study. The proposed testing methodology has been implemented in the R package HDTD.",
keywords = "Covariance matrix, high-dimensional settings, hypothesis testing, matrix-valued random variables, transposable data",
author = "Anestis Touloumis and Marioni, {John C.} and Simon Tavar{\'e}",
year = "2019",
month = "10",
day = "28",
doi = "10.5705/ss.202018.0268",
language = "English",
journal = "Statistica Sinica",
issn = "1017-0405",

}

Hypothesis Testing for the Covariance Matrix in High-Dimensional Transposable Data with Kronecker Product Dependence Structure. / Touloumis, Anestis; Marioni, John C.; Tavaré, Simon.

In: Statistica Sinica, 28.10.2019.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Hypothesis Testing for the Covariance Matrix in High-Dimensional Transposable Data with Kronecker Product Dependence Structure

AU - Touloumis, Anestis

AU - Marioni, John C.

AU - Tavaré, Simon

PY - 2019/10/28

Y1 - 2019/10/28

N2 - The matrix-variate normal distribution is a popular model for high-dimensional transposable data because it decomposes the dependence structure of the random matrix into the Kronecker product of two covariance matrices, one for each of the row and column variables. However, there is a lack of hypothesis testing procedures for the row or column covariance matrix in high-dimensional settings. Tests for assessing the sphericity, identity and diagonality hypothesis for the row (column) covariance matrix in high-dimensional settings while treating the column (row) dependence structure as a ‘nuisance’ parameter are introduced. The proposed tests are robust to normality departures provided that the Kronecker product dependence structure holds. In simulations, the proposed tests appear to maintain the nominal level and they tended to be powerful against the alternative hypotheses tested. The utility of the proposed tests is demonstrated by analyzing a microarray and an electroencephalogram study. The proposed testing methodology has been implemented in the R package HDTD.

AB - The matrix-variate normal distribution is a popular model for high-dimensional transposable data because it decomposes the dependence structure of the random matrix into the Kronecker product of two covariance matrices, one for each of the row and column variables. However, there is a lack of hypothesis testing procedures for the row or column covariance matrix in high-dimensional settings. Tests for assessing the sphericity, identity and diagonality hypothesis for the row (column) covariance matrix in high-dimensional settings while treating the column (row) dependence structure as a ‘nuisance’ parameter are introduced. The proposed tests are robust to normality departures provided that the Kronecker product dependence structure holds. In simulations, the proposed tests appear to maintain the nominal level and they tended to be powerful against the alternative hypotheses tested. The utility of the proposed tests is demonstrated by analyzing a microarray and an electroencephalogram study. The proposed testing methodology has been implemented in the R package HDTD.

KW - Covariance matrix

KW - high-dimensional settings

KW - hypothesis testing

KW - matrix-valued random variables

KW - transposable data

U2 - 10.5705/ss.202018.0268

DO - 10.5705/ss.202018.0268

M3 - Article

JO - Statistica Sinica

JF - Statistica Sinica

SN - 1017-0405

ER -