David Widmann

I think these questions touch [more general limitations of Random](https://github.com/JuliaRandom/RandomExtensions.jl) and of the `eltype` design - both which, as stated above, this PR intentionally does not want to address or...

> it sounds like this could vary case by case, and may not be expected to satisfy any particular invariant. Is that right? I expect that every `Distribution{

Personally, I think it would be fine to add Roots.jl as a dependency. It [only depends on Printf and CommonSolve](https://github.com/JuliaMath/Roots.jl/blob/master/Project.toml) which itself has [no dependencies](https://github.com/SciML/CommonSolve.jl/blob/master/Project.toml) and [only defines `solve` and...

I don't understand this PR anymore (even though I approved it :smile:). Wouldn't the "correct" fix be to use the existing `quantile_newton` etc. methods in https://github.com/JuliaStats/Distributions.jl/blob/ec8f7d82c59348443f799e0fb0394441fab31073/src/quantilealgs.jl#L37? And, if desired, add...

I am a bit unsure about this PR. I agree that it might be more convenient since one can skip an `using LinearAlgebra` or `using FillArrays` but it seems it...

I think we should qualify all functions, structs etc. from other packages in the docs, e.g., we should write `LinearAlgebra.I` or `LinearAlgebra.Diagonal` instead of `I` or `Diagonal` to make it...

What did you define that triggered the ambiguities? I would assume that you have to define different methods for the univariate case anyway since it operates on `Real`s whereas the...

The discussion in https://github.com/alan-turing-institute/MLJBase.jl/issues/504 seems related.

Ideally this should handle more general support types and not only multivariate or univariate variables if we generalize it.

They are tested with the standard tests in https://github.com/JuliaStats/Distributions.jl/blob/master/test/univariates.jl. See https://github.com/JuliaStats/Distributions.jl/blob/02bcbf8169160cc499832f05c83e9308ecdb2457/test/ref/continuous_test.ref.json#L1826 etc.