Data analysis for heavy tailed multivariate samples

Abstract

In this paper we propose methods of exploratory data analysis that test for joint stability of a multivariate data set. Examples are shown of how these techniques support or reject the joint stability of the sample. If these methods suggest that a data set is jointly stable, we would like to know how well methods of estimating a spectral measure work. We examine how two previously described estimators of the stable spectral measure perform under a variety of assumptions

Keywords:

• Multivariate stable distributions
• spectral measure
• estimation
• heavy tail modeling