Several standard texts (Anderson, Muirhead) state that for the rank deficient variant of a Wishart matrix, no density exists. However, it is always possible to build a rank deficient Wishart matrix generation routine and use it to evaluate empirical estimates of such a density (in the multivariate sense) over an ever increasing set of exemplars. To what does this sequence of empirical estimates converge in the limit of infinite exemplar count?

Similar questions and discussions