Making a combined CLs plots for various signal strength hypotheses

[ssaha1234]

Dear pyhf members,
I am creating a combination plot using CLs values across various hypotheses mass points for my analysis, I have made one CLs plot per mass point, something like this -

I would now like to make a plot, by combining all the masses I have (~12 such CLs plots above for 12 mass points), as in the plot here -

Is there a way to do this easily with cabinetry, for example ? I am looking at the example here, for reference - https://github.com/cabinetry/cabinetry-tutorials/blob/08a0363db0585d12bdcbc9600c9e139351142e8b/example.ipynb

Thanks !
Best,
Shreya

[alexander-held]

Hi Shreya, neither pyhf nor cabinetry currently have an interface for this kind of plot. It would be great to implement this somewhere centrally though (perhaps even mplhep could fit). I am sure that some people have already made that kind of plot with matplotlib, so perhaps we can build on existing implementations.

[alexander-held]

Related plots that might also be useful to implement: https://github.com/iris-hep/as-user-facing/blob/master/fit-visualization/exclusion-limits.md

[matthewfeickert]

@ssaha1234 We probably aren't going to prioritize this any time soon either on pyhf's side of things from the core dev team side as the visualization module is currently under contrib (pyhf.contrib.viz) and we would prefer to ideally have all visualization factored out into another library (eventually). That being said, the contrib name is supposed to be suggestive of the fact that we would be open to contributions from community members that want to implement this. 👍

Related plots that might also be useful to implement: https://github.com/iris-hep/as-user-facing/blob/master/fit-visualization/exclusion-limits.md

To emphasize Alex's point here, we do think these sorts of plots are quite important (we're users of them ourselves)! The primary point is that pyhf and cabinetry are libraries and not frameworks, so they should avoid scope creep that keeps them from achieving and improving on their primary goals, especially when other libraries in the Scikit-HEP ecosystem already specialize in visualization.

[ssaha1234]

Hi @alexander-held and @matthewfeickert,
Thanks a lot for pointing out to the scripts ! I can share the script I would be writing, once I have it working :)
Best,
Shreya

[kratsg]

I also have this notebook (https://gist.github.com/kratsg/4ff8cb2ded3b25552ff2f51cd6b854dc) which should help on some level.

[souravampire]

Hi @ssaha1234, did you solve it? I am stuck in exactly the same problem. I would appreciate it if you could share your progress.