MerQuR+Statistics Glossary

Statistics Glossary → D'Agostino's K² test

D'Agostino's K² testconcept

A normality test that assesses departure from the normal distribution by combining sample skewness and kurtosis.

The test first converts sample skewness and kurtosis into separate z scores based on their distributions under normality. It then adds the squares of the two: K² = Z₁² + Z₂². If the null hypothesis (the data are normal) holds, K² approximately follows a chi-square distribution with 2 degrees of freedom. It is also known as the D'Agostino-Pearson omnibus test.

It is useful because it also shows the kind of departure: one can see which component, skewness or kurtosis, is large. In very small samples (below roughly 20) the approximation for the kurtosis component weakens. Like every normality test, it flags practically trivial departures in large samples and can miss important ones in small samples, so it should be read alongside a Q-Q plot.

An everyday example

A laboratory checks whether the cholesterol readings of 200 patients are normally distributed. The skewness z score is 3.1 and the kurtosis z score 0.8, so K² = 9.61 + 0.64 = 10.25 and p is about 0.006. The departure comes mainly from right skew, meaning a few patients have very high cholesterol, while the tails are otherwise close to normal.

In MerQur

In the Normality Tests analysis, D'Agostino's K² is one of the seven tests under the 'Tests' checkboxes; if no test is ticked the choice is made automatically from n, and D'Agostino is also run when n >= 20.

Analyses where this concept appears

Normality Tests

Related concepts

SkewnessKurtosis