The skew elliptical distributions and their quadratic forms. (Q1426349)

Because of the increasing interest in using skew normal distributions in data fitting and analysis, the aim of the present paper is to investigate a family of skew elliptical distributions. In Section 2, the considered class of skew elliptical distributions is defined using the probability density function and specific stochastic representations, including skew normal distributions and a subset of elliptical distributions (spherical symmetric forms, invariant to orthogonal transformations). Section 3 deals with basic properties, including marginal and conditional distributions, distributions under linear transformations, and moments. In Section 4, the expression of the moment generating function of the skew elliptical distribution of \(k\) dimensions is obtained as an integral on the space of reduced dimension, equal to \(\min\{4, k+ 1\}\). Section 5 shows that, under certain conditions, the joint distribution of the quadratic forms of the skew elliptical distributions can be obtained. An analogue to the Fisher-Cochran theorem is proved to hold for skew elliptical distributions. An example of two-sided \(t\)-tests is given to illustrate how the distributions of some statistics can be derived as functions of quadratic forms.