MerQuR+Statistics Glossary

Statistics Glossary → Kurtosis

Kurtosisconcept

A shape measure based on the fourth standardized moment that shows how heavy or light a distribution's tails are compared with the normal distribution.

Kurtosis is the mean of the fourth powers of the standardized values; it equals 3 for the normal distribution. Most software subtracts 3 and reports excess kurtosis: 0 for the normal, positive for heavy-tailed (leptokurtic) distributions and negative for light-tailed (platykurtic) ones. Despite the popular description as 'peakedness', what it really measures is how often extreme values occur, not the shape of the peak.

Sample kurtosis is strongly affected by a few extreme values and is unstable in small samples, so it should not decide normality on its own but be read together with its standard error and with plots. Heavy tails can reduce the power of tests based on means and variances. Together with skewness, it underlies normality tests such as D'Agostino's K² and Jarque-Bera.

An everyday example

Delays on two bus routes are compared. Both have an average delay of 5 minutes and the same standard deviation. On the first route delays always stay between 3 and 7 minutes, while on the second most days have almost no delay and now and then a bus is 25 minutes late. The second route has clearly higher kurtosis.

In MerQur

As a shape measure (G2) together with its standard error in Descriptive Statistics and Likert Scale Analysis; in the descriptive tables of the t-tests, ANOVAs and nonparametric tests; and as Mardia's multivariate kurtosis test in MANOVA, Correlation Analysis, Canonical Correlation, Confirmatory Factor Analysis and Path Analysis.

Analyses where this concept appears

Descriptive StatisticsOne-Sample t-TestIndependent-Samples t-TestPaired t-TestOne-Way ANOVARepeated Measures ANOVAMANOVAMann-Whitney UWilcoxon Signed-RankKruskal-WallisFriedmanCorrelation AnalysisBland-Altman Agreement AnalysisCanonical CorrelationLikert Scale AnalysisConfirmatory Factor Analysis (CFA)Path Analysis

Related concepts

SkewnessNormality assumption