[PDF] Sensitivity of Normality Tests to Non-normal Data

What is Shapiro Wilk test?

THE SHAPIRO-WILK AND RELATED TESTS FOR NORMALITY GivenasampleX1,...,X nofnreal-valuedobservations, theShapiro– Wilk test (Shapiro and Wilk, 1965) is a test of the composite hypothesis that the data are i.i.d. (independent and identically distributed) and normal, i.e. N(µ,?2) for some unknown real µ and some ? > 0.

Are Shapiro-Wilk and W? asymptotically equivalent under the normality hypothesis?

The representations given by Leslie et al. are useful in comparing di?erent test statistics for normality. Namely, they show that under the normality hypothesis, the Shapiro–Wilk and Shapiro–Francia statistics W and W?are asymptotically equivalent in the sense (their equation (5)) that as n ? ?, n( ? W ? ? W?) ? 0 in probability.

What are the relative merits of the Shapiro-Wilk & Sha Piro-Francia tests?

The relative merits of the Shapiro–Wilk and Shapiro–Francia tests the versus skewness and kurtosis test have been a subject of debate. The interested reader is directed to the articles in the Stata Technical Bulletin.

How do you find the Shapiro Wilk statistic?

n) := m?V?1/C, which is a unit row vector. The Shapiro–Wilk statistic is then de?ned by (3) W = Xn j=1 a jX(j) !2 Xn j=1 (X j?X)2 ! , as in the paper of Shapiro and Wilk (1965). They point out that the statistic is preserved by a change in location, adding a constant b to all the X jand thus to each X(j).