clamp_errors options for Makie Errorbars and clamp_bincounts for histograms

[Moelf]

This is an alternative to #116 @sfranchel

I'm a bit hesitant to clip the error bars to positive

DO NOT MERGE

Blocked by MakieOrg/Makie.jl#3901

The following works on older Makie but not the latest one, I think it's a bug

julia> h = Hist1D(;binedges=0:1, bincounts=[0.1], sumw2=[0.1])
edges: [0.0, 1.0]
bin counts: [0.1]
total count: 0.1

julia> stephist(h); errorbars!(h); current_figure()

julia> stephist(h); errorbars!(h; error_function=FHist.pearson_err); current_figure()

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 46 lines in your changes are missing coverage. Please review.

Project coverage is 77.99%. Comparing base (e4bc8ca) to head (4a40c86).
Report is 1 commits behind head on main.

Files	Patch %	Lines
ext/FHistMakieExt.jl	0.00%	46 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #118      +/-   ##
==========================================
- Coverage   80.37%   77.99%   -2.39%     
==========================================
  Files          11       11              
  Lines         790      818      +28     
==========================================
+ Hits          635      638       +3     
- Misses        155      180      +25     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[sfranchel]

Thanks for looking into this!

I still think it would be very useful to clip both negative/zero values and errors to positive, to do plots in log scale. I'm quite frequently getting into errors for non-compatible values for log scale. It's true that one could avoid it by choosing more wisely the binning of histograms, but when you need to run on high number of events and a bit more complex selections it isn't really obvious what's the best binning, and it's very likely to end up with empty bins. For several MC generators, it's also quite common to get negative weights, which can produce negative valued bins.

So I'd love to have an option to clip bin counts and errors! If you believe this is more of an edge case, we could set it as not being the default behaviour.

[Moelf]

Negative bin counts tell you something is happening with the MC, and users should take conscious decisions about it, we don't want to silently produce incorrect visualization is my take.

I'm happy to clamp error bars automatically

[sfranchel]

Totally agree that we shouldn't silently produce incorrect visualization, indeed I think it would bet better to not have on this feature by default.

[Moelf]

stephist(h); errorbars!(h; clamp=false); current_figure()

julia> stephist(h); errorbars!(h); current_figure()

julia> stephist(h); errorbars!(h; error_function=FHist.pearson_err); current_figure()

@sfranchel how about this?

[sfranchel]

Ok, that looks good. Maybe I'd make the default clamp=false, and I'd add the same option for clamping also bin counts in barplots and scatter plots 🙏

[Moelf]

ah, right now it only clamps the error bars -- since it's a keyword argument to the errorbars(), I thought we want to clamp error bars above zero by default, we just didn't want to clamp the bincounts automagically.

Actually, I think for the bincounts(), we should just add clamp() method to histograms themselves?

[Moelf]

julia> h = Hist1D(;binedges=0:2, bincounts=[-0.1, 0.1], sumw2=[0.1, 0.1])
edges: [0.0, 1.0, 2.0]
bin counts: [-0.1, 0.1]
total count: 0.0

julia> clamp_bincounts = true; clamp_errors=true;

julia> stairs(h; clamp_bincounts); errorbars!(h; clamp_bincounts, clamp_errors); current_figure()

Why the default behavior?

• It's important for people to catch real negative bin counts -- it's an indication of low stats of generated events
• It's more annoying than helpful to deal with error bars that happen to dip under zero most of the time.

-	clamp_bincounts = false (default)	clamp_bincounts = true
clamp_errors = false		Don't do this
clamp_errors = true (default)

[Moelf]

one caveat is that until MakieOrg/Makie.jl#3904 is addressed, hist() and stephist() won't be able to use clamp_bincounts = true, you need to directly use barplot() and stairs(). They are the same, but it is a bit annoying

[Moelf]

julia> h = Hist1D(;binedges=0:2, bincounts=[-0.1, 0.1], sumw2=[0.1, 0.1])
edges: [0.0, 1.0, 2.0]
bin counts: [-0.1, 0.1]
total count: 0.0

julia> clamp_bincounts = true; clamp_errors=true;

julia> barplot(h; clamp_bincounts); errorbars!(h; clamp_bincounts, clamp_errors); current_figure()

-	clamp_bincounts = false (default)	clamp_bincounts = true
clamp_errors = false		Don't do this
clamp_errors = true (default)

[Moelf]

for completeness: scatter plot:

[sfranchel]

Thanks a lot @Moelf , this looks great!! Very much appreciated