pyhf / pyhf-validation Goto Github PK

While doing the validation for the displaced slepton analysis, it was found that it would be helpful to fix the mu_Sig parameter to 0 to do a bkg-only fit when comparing the fitted nuisance params in compare_fitted_nuisance.py . To do so, the following mods were needed:

line 58: bestfit = pyhf.infer.mle.fit(data, model) --> bestfit = pyhf.infer.mle.fixed_poi_fit(0.0, data, model)

After line 25: workspace.var('mu_Sig').setVal(0.0)

One of the steps for validating pyhf against ROOT is comparing nuisance parameters, which unfortunately involves some translation of the NP names.

To that end, it looks like one of these duplicated lines should be for "shape" systematics.

https://github.com/pyhf/pyhf-validation/blob/master/scripts/compare_nuisance.py#L37-L38

In my specific use case, this did the trick (but not sure how robust this is for other workspaces):
.replace("gamma_shape_", "shape_")

Transferring some information and ideas from a developer chat with @kratsg:

• Support CLI that will allow for

pyhf-validation spec.json

that will run pyhf cls over spec.json and additionally runs json2xml+hist2workspace+run_single.py on the resulting ROOT workspace from the HistFactory side.

• Provide signatures of

delta, pyhf_out, hifa_out = pyhf-validation.run_validation("spec.json")

• Support ROOT and PyROOT in automated testing
  • Maybe treat it as a soft dependency (try import ROOT; except foo)

In the CI

will often fail as it will take longer than 1 second for test_version to complete but only on MacOS. I am not sure why it is only MacOS, but this should be resolved.

c.f. PR #9