Testing for normality in arbitrary dimension (Q1820530)

\textit{K. Murota} and \textit{K. Takeuchi} [Biometrika 68, 55-65 (1981; Zbl 0463.62041)] proposed a test for univariate normality based on the studentized empirical characteristic function. The present paper uses a similar technique to test the hypothesis that a multivariate distribution is a normal distribution. The studentized multivariate empirical characteristic function is shown to converge weakly (as the sample size increases) to a multivariate Gaussian process. The distribution of the absolute supremum of this Gaussian process is studied. The test is based on the maximal deviation of the absolute value of the studentized empirical characteristic function from the absolute value of a normal characteristic function, and is shown to be consistent against a class of alternatives to normality.