David Widmann

Sorry, I missed the last few comments here. I added more tests (it seems currently `pairwise!` with `dims` is not tested). I was not sure if the plan is to...

It depends :slightly_smiling_face: MvNormal works with covariance matrices of type `PDMats.AbstractPDMat` from [PDMats.jl](https://github.com/JuliaStats/PDMats.jl). You can provide such a positive definite matrix type explicitly when constructing an `MvNormal` distribution, otherwise the...

And here's the implementation in Distributions: https://github.com/JuliaStats/Distributions.jl/blob/ed52948fb47e7e0aecaf81e99ebb090c2738cf0e/src/multivariate/mvnormal.jl#L258-L267

I would be open to it. I don't agree with the sentiment in the original issue that it is a Bayesian "thing" and therefore should not be added - I...

A PR would be appreciated. > it's also not clear to me what `α0` is supposed to be--the usual parametrization of a `DirichletMultinomial` is just in terms of `n` and...

Indeed. I guess, however, here we would want the definitions suggested by @sethaxen in any case, even if we could dispatch on such traits.

@oxinabox maybe you know if this would have fixed the problems you had?

Hmm these errors wouldn't be fixed by the PR unfortunately. I guess we can hold back this PR until some specific use case pops up that would be solved by...

https://github.com/JuliaStats/Distributions.jl/issues/1602#issuecomment-1209901969 explains how one can use positive semi-definite matrices (to some extent, i.e., e.g. for `rand`).

> @bicycle1885 are you still interested in finishing this PR? @devmotion made a reasonable code suggestion that you would just need to accept, the other thing that seems missing is...