Talk:Matrix t-distribution

Move discussion in progress

There is a move discussion in progress on Talk:Student's t-distribution which affects this page. Please participate on that page and not in this talk page section. Thank you. —RMCD bot 04:01, 24 August 2021 (UTC)[reply]

Generalized(?) matrix t distribution

The "generalized" version of the distribution is actually just a re-parameterization, not an extension of the distribution family defined by the "standard" formulation. The "generalized" form is based on the Inverse Matrix Gamma distribution, which is also just a re-parameterization of the inverse Wishart, just like the Matrix Gamma is another formulation of the Wishart distribution. I suggest we should rename "Generalised" -> "Re-parameterized", and also point out that it is over-parameterised. ArneLeijon (talk) 13:06, 1 July 2024 (UTC)[reply]

Conditions for defined Mean and Covariance

Dear previous contributors, Please take a look at my latest edits. I have added conditions for well-defined Mean and Covariance to exist, to be consistent with the (vector-)multivariate t-distribution. I am sure the previous version had an incorrect condition for the Mean in the box for the "standard" Matrix t version.

Gupta & Nagar (2000, Theorem 4.3.1) stated the mean as E[T]= M without any restriction on the degree of freedom parameter, and Iranmanesh et al (2010) states the same with ref to Gupta. However, I still think there are good reasons to keep the statement for the Matrix t consistent with the corresponding condition for the vector-multivariate t-distribution: Since the Matrix t distribution is equivalent to the vector-multivariate version if the matrix has only one row, or only one column, the requirements on the degree of freedom parameter must be the same, and I am sure the article on the (vector-)multivariate t distribution is correct on this point. ArneLeijon (talk) 20:29, 1 July 2024 (UTC)[reply]