Modification of Anderson-Darling goodness-of-fit test for normality

Abstract

The objective of the paper is to present a new goodness-of-fit test for normality. This new proposition, here called the ADm test, is a modification of the Anderson-Darling goodness-of-fit test for normality. Critical values of the ADm test are determined by a
Monte Carlo simulation for sample sizes. The ADm test was compared with the other normality tests with respect to their power. The simulation study shows that the ADm test is the most powerful for symmetric short tailed distributions and for symmetric distributions close to the normal distribution. The test has also the best performance for asymmetric short tailed distributions.