This is a cross-posting of a question I asked at CrossValidated. It hasn't generated much activity so I'm trying here:

Suppose $X\sim \operatorname{InvWishart}(\nu, \Sigma_0)$. I'm interested in the marginal distribution of the diagonal elements $\operatorname{diag}(X) = (x_{11}, \dots, x_{pp})$. There are a few simple results on the distribution of submatrices of $X$. From these I can figure that the marginal distribution of any single element on the diagonal is inverse gamma. But I've been unable to deduce the joint distribution. I suspect that I'm missing something simple; it seems like this "ought" to be known but I haven't been able to find/show it.

  • $\begingroup$ I'd have thought this would be in the literature somewhere. It certainly is for the Wishart distribution (the diagonal entries are chi-square-distributed, IIRC). $\endgroup$ – Michael Hardy Sep 7 '11 at 22:09
  • $\begingroup$ Hi Michael, do you have any progress on this topic? I had the same question, and don't want to start a new one, so just want to ask if you have any reference or maybe answer on this question. As you said, the diagonal entries for Wishart distribution are chi-square distributed, and the y are inverse-gamma distributed for Inverse Wishart. Do you know the joint marginal distribution? $\endgroup$ – MPQ Sep 22 '11 at 19:18

In the following reference

\textsc{Letac, G. and Weso{\l}owski, J.} (2000) 'An independence property for the product of GIG and gamma laws.' {\it Ann. Probab.} {\bf 28}, 1371-1383.

you will find at the end of the paper some remarks about the function $K(a,b,p)=\int_{P_n}\exp( -trace[ax+bx^{-1}])\det^{p-(n+1)/2}dx$ where $P_n$ is the cone of positive definite matrices where $a,b\in P_n$ and where $p>(n-1)/2.$ Up to some constants, the function $b\mapsto K(a,b,p)$ is basically the Laplace transform of $X^{-1}$ when $X$ is Wishart with parameters $p$ and $a.$ Taking $b$ as a diagonal matrix gives you the Laplace transform of the diagonal elements of the random matrix $X^{-1}.$


See http://www.math.yorku.ca/~massamh/sjos_391.pdf It appears as a solution is given in this paper


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.