How to use tests for univariate normality to assess multivariate normality

Research output: Contribution to journalArticle

64 Citations (Scopus)

Abstract

The assumption of multivariate normality (MVN) underlies many important techniques in multivariate analysis. In the past 50 years, over 50 tests of this assumption have been proposed. However, for various reasons, practitioners are often reluctant to address the MVN issue. In this article, several techniques for assessing MVN based on well-known tests for univariate normality are described and suggestions are offered for their practical application. The techniques are illustrated using two previously published sets of real-life data. In one of the examples it is shown that simply testing each of the marginal distributions for univariate normality can lead to a mistaken conclusion.

Original languageEnglish (US)
Pages (from-to)64-70
Number of pages7
JournalAmerican Statistician
Volume49
Issue number1
DOIs
StatePublished - Jan 1 1995
Externally publishedYes

Fingerprint

Multivariate Normality
Normality
Univariate
Multivariate Analysis
Marginal Distribution
Testing

Keywords

  • Box-Cox procedure
  • Fisher iris data
  • Kurtosis
  • Shapiro-Wilk test
  • Skewness

ASJC Scopus subject areas

  • Statistics and Probability
  • Mathematics(all)
  • Statistics, Probability and Uncertainty

Cite this

How to use tests for univariate normality to assess multivariate normality. / Looney, Stephen Warwick.

In: American Statistician, Vol. 49, No. 1, 01.01.1995, p. 64-70.

Research output: Contribution to journalArticle

@article{7c7c8743f688464987448f960f105f35,
title = "How to use tests for univariate normality to assess multivariate normality",
abstract = "The assumption of multivariate normality (MVN) underlies many important techniques in multivariate analysis. In the past 50 years, over 50 tests of this assumption have been proposed. However, for various reasons, practitioners are often reluctant to address the MVN issue. In this article, several techniques for assessing MVN based on well-known tests for univariate normality are described and suggestions are offered for their practical application. The techniques are illustrated using two previously published sets of real-life data. In one of the examples it is shown that simply testing each of the marginal distributions for univariate normality can lead to a mistaken conclusion.",
keywords = "Box-Cox procedure, Fisher iris data, Kurtosis, Shapiro-Wilk test, Skewness",
author = "Looney, {Stephen Warwick}",
year = "1995",
month = "1",
day = "1",
doi = "10.1080/00031305.1995.10476117",
language = "English (US)",
volume = "49",
pages = "64--70",
journal = "American Statistician",
issn = "0003-1305",
publisher = "American Statistical Association",
number = "1",

}

TY - JOUR

T1 - How to use tests for univariate normality to assess multivariate normality

AU - Looney, Stephen Warwick

PY - 1995/1/1

Y1 - 1995/1/1

N2 - The assumption of multivariate normality (MVN) underlies many important techniques in multivariate analysis. In the past 50 years, over 50 tests of this assumption have been proposed. However, for various reasons, practitioners are often reluctant to address the MVN issue. In this article, several techniques for assessing MVN based on well-known tests for univariate normality are described and suggestions are offered for their practical application. The techniques are illustrated using two previously published sets of real-life data. In one of the examples it is shown that simply testing each of the marginal distributions for univariate normality can lead to a mistaken conclusion.

AB - The assumption of multivariate normality (MVN) underlies many important techniques in multivariate analysis. In the past 50 years, over 50 tests of this assumption have been proposed. However, for various reasons, practitioners are often reluctant to address the MVN issue. In this article, several techniques for assessing MVN based on well-known tests for univariate normality are described and suggestions are offered for their practical application. The techniques are illustrated using two previously published sets of real-life data. In one of the examples it is shown that simply testing each of the marginal distributions for univariate normality can lead to a mistaken conclusion.

KW - Box-Cox procedure

KW - Fisher iris data

KW - Kurtosis

KW - Shapiro-Wilk test

KW - Skewness

UR - http://www.scopus.com/inward/record.url?scp=21844509888&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=21844509888&partnerID=8YFLogxK

U2 - 10.1080/00031305.1995.10476117

DO - 10.1080/00031305.1995.10476117

M3 - Article

VL - 49

SP - 64

EP - 70

JO - American Statistician

JF - American Statistician

SN - 0003-1305

IS - 1

ER -