pdeffebach

> Actually now I have checked the `methods(cov)` and `methods(cor)` - maybe for higher dimensional arrays we can just allow `dims` argument as for 2-dimensional case? This is a potential...

> (so I feel it is safer to exclude `AbstractArray` for now other than vectors and matrices). Current implementation excludes `cov(Any, Matrix)` as well as other higher dimensional arrays. >...

> Regarding `skipmissings` - yes, I have just noticed it was added but not released yet. But how do you plan to handle "pairwise complete observations" with this design cleanly?...

There is currently no `skipmissing` implementation that preserves the dimensionality of an array. So that is an open problem that will have to be solved by additions to `skipmissing` or...

Thank you for your comments. Here are my thoughts: I think it's important to understand the purpose `skipmissing` serves. A researcher gets a data-set and wites functions on a subset...

Thanks for your hard work on this. I know what a tricky PR it is, I really think something like [`passmissing`](https://github.com/JuliaData/DataFramesMeta.jl/pull/276) for vectors is the solution. We don't have to...

> Are you sure? I'm not saying we should keep dispatch based on `AbstractWeights`, but given that it's an `AbstractVector` it should work well with `passmissing_vec`, whether it's passed as...

> * Use a more complex design with a macro and escape sequences to specify which arguments should not be modified, like `@passmissing quantile(x, $p)` We could also do `passmissing(quantile)(x,...

> Now the challenge is that general Tables.jl tables will not be able to support `passmissing_vec` as the interface does not specify a way to drop observations, so we would...

In an ideal world, this goes in the model. The ideal model is something like ``` formula = @formula(y ~ x | fe = absorb(x, t) | cluster(state) | dropmissing(all)...