In Monte Carlo simulation of multivariate distributions, it is often helpful to use a general class of distributions which share certain defining characteristics but which allow controlled variation of other characteristics. We show how multivariate kurtosis, as measured by Mardias coefficient, Beta 2,p, can be controlled across the class of elliptically-contoured distributions. This allows convenient assessment of the effects of kurtosis on test power, robustness, or whatever the Monte Carlo subject of interest. We illustrate the methods utility by showing that common tests for skewness are also very sensitive to kurtosis even in nonskewed distributions.
This article is from 'Computing Science and Statistics: Proceedings of the Symposium on Interface Critical Applications of Scientific Computing (23rd): Biology, Engineering, Medicine, Speech Held in Seattle, Washington on 21-24 April 1991,' AD-A252 938, p466-469.