Sub-dimensional Mardia measures of multivariate skewness and kurtosis
Abstract
The Mardia measures of multivariate skewness and kurtosis summarize the respective characteristics of a multivariate distribution with two numbers. However, these measures do not reflect the sub-dimensional features of the distribution. Consequently, testing procedures based on these measures may fail to detect skewness or kurtosis present in a sub-dimension of the multivariate distribution. We introduce sub-dimensional Mardia measures of multivariate skewness and kurtosis, and investigate the information they convey about all sub-dimensional distributions of some symmetric and skewed families of multivariate distributions. The maxima of the sub-dimensional Mardia measures of multivariate skewness and kurtosis are considered, as these reflect the maximum skewness and kurtosis present in the distribution, and also allow us to identify the sub-dimension bearing the highest skewness and kurtosis. Asymptotic distributions of the vectors of sub-dimensional Mardia measures of multivariate skewness and kurtosis are derived, based on which testing procedures for the presence of skewness and of deviation from Gaussian kurtosis are developed. The performances of these tests are compared with some existing tests in the literature on simulated and real datasets.
Related Papers
- → Nonparametric MANOVA Approaches for Non-Normal Multivariate Outcomes(2013)15 cited
- → A Simulation Comparison of Univariate and Multivariate Analyses of a Multi-Factor Repeated Measures Design.(1981)2 cited
- → Bivariate and multivariate data(1989)1 cited
- A suggestion for a multivariate concordance coefficient(2014)
- Regression of Random Vectors for Multivariate t-Student Distribution(2011)