Allow weights in statistical recipes

[greimel]

Add the keyword weights to boxplot, density, hist, violin.

density

let
	N = 100_000
	x = rand(Uniform(-2, 2), N)
	
	w = pdf.(Normal(), x)

	fig = Figure()
	density(fig[1,1], x)
	density(fig[1,2], x, weights = w)

	fig
end

violin

let
	N = 100_000
	x = rand(1:3, N)
	y = rand(Uniform(-1, 5), N)
	
	w = pdf.(Normal(), x .- y)

	fig = Figure()

	violin(fig[1,1], x, y)
	violin(fig[1,2], x, y, weights = w)

	fig
end

boxplot

let
	N = 100_000
	x = rand(1:3, N)
	y = rand(Uniform(-1, 5), N)
	
	w = pdf.(Normal(), x .- y)

	fig = Figure()

	boxplot(fig[1,1], x, y)
	boxplot(fig[1,2], x, y, weights = w)

	fig
end

hist

let
	N = 100_000
	x = rand(Uniform(-5, 5), N)
	
	w = pdf.(Normal(), x)

	fig = Figure()
	hist(fig[1,1], x)
	hist(fig[1,2], x, weights = w)

	fig
end

[greimel]

Friendly bump 😃

The added functionality generalizes what AlgebraOfGraphics.density() and AlgebraOfGraphics.histogram() can already do. (e.g. data(df) * mapping(:y, weights = :wgts) * density() |> draw)

@piever, since you have written most of the code that this PR touches – would you mind reviewing it if you find some time?

[piever]

We should probably add a utility function to filter out automatic keywords, but that should be done separately.

[piever]

Nice! Looks good to me (just left two tiny formatting comments).

I have a minor doubt as to whether we should call StatsBase.weights indiscriminately, or instead not call it when the user passes a StatsBase.AbstractWeight object. I have no idea whether the various subtypes of AbstractWeight make any difference here. I suspect not, so probably it's fine to leave as is.

[greimel]

I've applied your suggestions, thanks for the review!

[SimonDanisch]

Great!
There is still an error in the docs:

┌ Franklin Warning: in <examples/plotting_functions/violin.md>
│ There was an error of type 'ArgumentError' when running a code block.
│ Checking the output files '/home/runner/work/Makie.jl/Makie.jl/docs/__site/assets/examples/plotting_functions/violin/code/output/example_figure.(out|res)'
│ might be helpful to understand and solve the issue.
│ 
│ Relevant pointers:
│ - evaluation of Julia code blocks: https://franklinjl.org/code/
│ 
│ Details:
│ ArgumentError: Package Distributions not found in current path:
│ - Run `import Pkg; Pkg.add("Distributions")` to install the Distributions package.

Can you add that to the docs project?

[greimel]

Done! I hadn't realized that the docs are actually built for non-member contributors these days. That's great news!

[greimel]

I would like to use this for a lecture in about a week.

If it's not too much of a hassle for you, I'd appreciate this being merged and tagged before May 8.

If you don't plan a new tag soon, please let me know and I'll adjust my material accordingly

[SimonDanisch]

I'll try to release a new version!

[SimonDanisch]

Would a breaking version work for you as well?

[greimel]

that's fine, thanks for asking!