normalization = : pdf argument in hist plot does not normalize correctly

[danielchen26]

• are you running newest version (version from docs) ?

  • yes

  • (temp) pkg> st
    status `~/documents/temp/project.toml`
      [13f3f980] cairomakie v0.11.9
      [ee78f7c6] makie v0.20.8
    

• can you reproduce the bug with a fresh environment ? (]activate --temp; add Makie)

  • yes

• What platform + GPU are you on?

  • mac

issues for the following code:

using CairoMakie
testd = 0.5 * randn(1000) .+ 3.5
testd2 = 0.2 * randn(1000) .+ 5.5

hist(testd .+ testd2, normalization = :pdf, bar_labels = :values,
     label_formatter=x-> round(x, digits=2), label_size = 15,
     strokewidth = 0.5, strokecolor = (:black, 0.5), color = :values)

The normalization is wrong here.

[asinghvi17]

This is a StatsBase issue and should probably be raised there. Consider the following:

julia> h = StatsBase.fit(StatsBase.Histogram, testd .+ testd2)
Histogram{Int64, 1, Tuple{StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}, Int64}}}
edges:
  7.0:0.5:11.0
weights: [6, 34, 158, 326, 315, 131, 27, 3]
closed: left
isdensity: false

julia> h_norm = StatsBase.normalize(h; mode = :pdf)
Histogram{Float64, 1, Tuple{StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}, Int64}}}
edges:
  7.0:0.5:11.0
weights: [0.012, 0.068, 0.316, 0.652, 0.63, 0.262, 0.054, 0.006]
closed: left
isdensity: true

julia> h_norm = StatsBase.normalize(h; mode = :probability)
Histogram{Float64, 1, Tuple{StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}, Int64}}}
edges:
  7.0:0.5:11.0
weights: [0.006, 0.034, 0.158, 0.326, 0.315, 0.131, 0.027, 0.003]
closed: left
isdensity: false

So you will probably get what you're looking for using normalization = :probability.

I'll close this issue for now, since it's not a problem we can solve in Makie directly.