add varm handling for Missing

[bkamins]

Now std([1, 2, missing]) fails, because varm requires that the second argument is <:Number. As this is a basic functionality in Base I think it would be good to add a custom method handling Missing in Missings.jl.

There are similar failures in many functions in StatsBase, but I guess that this should be handled elswhere? Yes?.

[nalimilan]

You're supposed to do std(skipmissing([1, 2, missing])) in that case. That's documented in the README and in the DataFrames manual, hopefully once Missing is added to Base we'll add a section in the manual about it.

It would make sense to have std accept any AbstractArray{<:Union{Number, Missing}}) input, and return missing if there's a missing value, but that will only be workable with Missing defined in Base: else we would have to duplicate all math functions taking arrays.

[bkamins]

Thank you for the answer.

It would make sense to have std accept any AbstractArray{<:Union{Number, Missing}}) input, and return missing if there's a missing value

This is exactly what I have meant (I know skipmissing, but it is a different use case).

Close this issue then?

[nalimilan]

I guess we can keep it open as a reminder of what needs to be done in Base.