Comparison of selected tests for univariate normality based on measures of moments

Czesław Domański, Piotr Szczepocki

Univariate normality tests are typically classified into tests based on empirical distribution, moments, regression and correlation, and other. In this paper, power comparisons of nine normality tests based on measures of moments via the Monte Carlo simulations is extensively examined. The effects on power of the sample size, significance level, and on a number of alternative distributions are investigated. None of the considered tests proved uniformly most powerful for all types of alternative distributions. However, the most powerful tests for different shape departures from normality (symmetric short-tailed, symmetric long-tailed or asymmetric) are indicated.

MoreLess

Year of publication:	2020
Authors:	Domański, Czesław ; Szczepocki, Piotr
Published in:	Statistics in transition : an international journal of the Polish Statistical Association and Statistics Poland. - Warszawa : GUS, ISSN 2450-0291, ZDB-ID 2235641-1. - Vol. 21.2020, 5, p. 151-178
Subject:	normality tests \| Monte Carlo simulation \| power of test \| Monte-Carlo-Simulation \| Statistischer Test \| Statistical test \| Theorie \| Theory \| Simulation