Some properties of complex matrix-variate generalized Dirichlet integrals (Q2572420)

Let \(X_1,\dots,X_k\) be hermitian positive definite \(p\times p\) matrix random variables having the joint density function \[ \begin{multlined} f(X_1,\dots,X_k)= c_k|\det(X_1)|^{\alpha_1-p}\dots|\det(X_p)|^{\alpha_p-p} \\ \times|\det(I+X_2+\dots+X_k)|^{\beta_1}\dots|\det(I+X_k)|^{\beta_{k-1}} \\ \times|\det(I+X_2+\dots+X_k)|^{-(\alpha_1+\dots+\alpha_{k+1}+\beta_1+\beta_k)}\;\;\text{for }\text{Re}(\alpha_j)>p-1 \end{multlined}\tag{\(*\)} \] \(f(X_1,\dots,X_k)=0\quad \text{elsewhere}\), where \(c_k\) is the normalizing constant. The authors derive the value of \(c_k\) and establish five theorems for the above defined probability distribution. One of the theorems established by the authors is given below. Theorem 2.1. Let \(X_1,\dots,X_k\) be matrix-variate random variables having the joint density as given above in (\(*\)). Then for the transformation \[ \begin{aligned} Y_1&=(I+X_2+\dots+X_k)^{-\frac12} X_1(I+X_2+\dots+X_k)^{-\frac12}\\ &\;\vdots\\ Y_k&=(I+X_k)^{-\frac12}(I+X_k)^{-\frac12},\end{aligned} \] \(Y_1,\dots,Y_k\) are independent, and further, \(Y_j\) has type-1 density with parameters \((\alpha_j,\alpha_{j+1}+\dots+\alpha_{k+1}+\beta_j+\beta_k)\), \(j=1,\dots,k\).