Carlos Parada

> For analytic weights the situation is more tricky as the definition isn't completely clear in Stata nor in StatsBase (see [JuliaStats/StatsBase.jl#758](https://github.com/JuliaStats/StatsBase.jl/issues/758)). But it seems that at least when all...

> Great point @ParadaCarleton , and thanks for the link! > > A `μ,ν` parameterization for the Beta is a great idea, and `StudentT` also uses `ν` for degrees of...

> This definition sounds fine as it's quite precise. Though as long as the estimator makes some sense, I wouldn't completely deprecate it, but instead add a warning to the...

> the latest version handles this transform better btw but there are still odd edge cases where it's not semantically equivalent code. In that case, maybe this should be off...

> That would be a great feature I think. I am not very familiar with weighted samples so am not sure what a good design choice would be. > >...

It has now been 2 years and I have no idea what I was talking about then :smile: But I think [this](https://link.springer.com/content/pdf/10.1007/978-3-319-23528-8_5.pdf) was the relevant paper that inspired this suggestion,...

Briefly going to comment on this to say--the solution to this issue is to use ReverseDiff or ForwardDiff.jl or (a few years down the line when it's mature) maybe some...

> There are some internals of Turing that will need to be made compatible with the Cholesky type. One known example is packaging samples into MCMCChains. If we switch to...

I think I mentioned elsewhere that I'm very much against VIF being added to any packages, as it's a statistical footgun in the waiting. I'll explain why here. The VIF...

> The typical applications I've seen in my domains is an indicator of whether or not individual coefficient standard errors are conflated or whether there are potential numerical issues. I...