Functions of Matrix Argument, Scaling Models and Symmetric Product of Matrices

A real-valued scalar function f(X) will be called a density, if f(X) ≥ 0 for all X in the domain of X, and X f(X)dX = 1, where dX represents the wedge product of the differentials of all distinct real scalar varaiables in X, with X being a scalar (a 1 × 1 matrix)/vector (a p ×1 or 1×p matrix)/matrix variable in the real or complex domain. If U =√aX,a>0, then we say that X is scaled by a real scalar quantity. If U = A1 2X,A > O, then we say that X is scaled by a real positive definite matrix when the product is defined. This paper deals with the cases when X is a scalar/vector/matrix in the real or complex domain having a very general density f(X) belonging to Mathai’s pathway family, namely matrix-variate gamma, type-1 beta and type-2 beta distributions and the scaling factor a > 0 or A > O. Explicit forms of the densities for all different combinations, in the real and complex domains, are given. Their connections to Bayesian analysis and the case, when a or A and X are statistically independent, is also considered. An idea of rectangular matrix-variate distributions for the scaling factor A is explored. Some open problems are also pointed out.

View article page

Download PDF