David Widmann
David Widmann
Thank you for the PR! I wanted to add these improvements, so I applied the suggestions and hope this is fine with you.
I reviewed the PR and it seems the PR is clearly not in a mergeable state. The comments have to be addressed first but apparently this has not been done.
No, we don't use a specific style here (yet). (BTW in general running the JuliaFormatter won't necessarily revert all of your changes.)
The `eltype` definition is incorrect. It should return the element type of the samples for arrayvariate distributions (ie., e.g., `Float64` if the samples are `AbstractVector{Float64}`) but currently in the multivariate...
AFAICT `fit_map` was moved to https://github.com/JuliaStats/ConjugatePriors.jl (which would also explain the broken link).
Hmm yes it seems it has to be updated a bit (which seems totally possible though, I left some comments in @mschauer's PR). I think ideally we would point to...
Some more high-level comments: - I think we should not introduce any additional API. and hence in particular not add `affine`: We should continue to use `*` and `+` (and...
> I agree we should encourage and use `*`, `+`, `/`, and `-` over `affine`, but I think it makes sense to include `affine` for ease of implementation. This way...
No, this is not sufficient. We want to create and work with `Distribution`s, `*` etc. should not suddenly lift `Distribution`s to `Measure`s.
We have to support `+`, `*` etc. anyway if we want to avoid a breaking change since it is the current API. To me it seems an additional function would...