Testing normality in any dimension by Fourier methods in a multivariate Stein equation

TL;DR

This paper introduces a new affine invariant test for multivariate normality based on a Stein equation characterization, with proven limit distributions and strong simulation performance.

Contribution

It develops a novel, consistent, affine invariant test for multivariate normality using a Stein equation approach, including limit distribution derivations and practical performance evaluation.

Findings

Test has strong power compared to competitors.

Asymptotic confidence intervals are accurate.

Method performs well in real data example.

Abstract

We study a novel class of affine invariant and consistent tests for multivariate normality. The tests are based on a characterization of the standard $d$-variate normal distribution by means of the unique solution of an initial value problem connected to a partial differential equation, which is motivated by a multivariate Stein equation. The test criterion is a suitably weighted $L^2$-statistic. We derive the limit distribution of the test statistic under the null hypothesis as well as under contiguous and fixed alternatives to normality. A consistent estimator of the limiting variance under fixed alternatives as well as an asymptotic confidence interval of the distance of an underlying alternative with respect to the multivariate normal law is derived. In simulation studies, we show that the tests are strong in comparison with prominent competitors, and that the empirical coverage rate of the asymptotic confidence interval converges to the nominal level. We present a real data example, and we outline topics for further research.