An affine invariant multiple test procedure for assessing multivariate normality

: A multiple test procedure for assessing multivariate normality (MVN) that combines a finite set of affine invariant test statistics for MVN is proposed. This combination is based on a method introduced by Fromont and Laurent (Ann. Statist., 680–720, 34, 2006) that can be viewed as an improvement of the classical Bonferroni’s method. The usefulness of such approach is illustrated through a multiple test involving Mardia’s and BHEP (Baringhaus-Henze-EppsPulley) tests that are among the most recommended procedures to test a MVN hypothesis. A simulation study carried out for a wide range of alternative distributions to analyze the finite sample power behaviour of the proposed multiple test procedure indicates that the new test presents a good overall performance against other highly recommended MVN tests.

• Type: Journal
• Year: 2011
• Where: CSDA
• Authors: Carlos Tenreiro
• Keywords: CSDA 2011, Invariant Test Statistics, Multiple Test, Multiple Test Procedure, Social Sciences