Association and infinite divisibility for the Wishart distribution and its diagonal marginals (Q2640281)

Let S be a measurable space with a partial order. An S-valued random variable X is said to be associated if Cov[f(x),g(x)]\(\geq 0\) for every pair of bounded, measurable functions \(f,g: S\to R\) that are each increasing in the partial order. In the case \(S={\mathbb{R}}^ n\), \[ (x_ 1,...,x_ n)\leq (y_ 1,...,y_ n):\Leftrightarrow \quad x_ i\leq y_ i,\quad 1\leq i\leq n. \] If S is the additive semigroup of nonnegative definite matrices then \[ A\leq B:\Leftrightarrow \quad B-A\in S,\quad A,B\in S. \] The author gives an example of a central Wishart matrix W with one degree of freedom and scale matrix of rank 2 such that the diagonal entries of W are not associated. This allows him to conclude that no central Wishart matrix with one degree of freedom and scale matrix of rank 1 is associated. Employing the connection between association and infinite divisibility he shows that there exist Gaussian vectors such that the vector of squares is not infinitely divisible and obtains another proof of Lévy's result that no central Wishart matrix with scale matrix of rank greater than 1 is infinitely divisible.