iris-hep/analysis-grand-challenge

Port histogram re-binning to `main`

alexander-held opened this issue · 4 comments

alexander-held commented

Port over #165 (and eventually replace by a cabinetry-native method) to main after #158 is merged.

The lower bin edge shift from 120j -> 110j should not only be done in the histogram saving but also in the notebook visualization.

alexander-held commented

In the context of this update (which will change histograms) it would be useful to also fix the inconsistent use of >= / > for b-tagging cuts as pointed out by @eguiraud:

analysis-grand-challenge/analyses/cms-open-data-ttbar/ttbar_analysis_pipeline.py

Lines 254 to 255 in ca06789

selections.add("exactly_1b", ak.sum(jets.btagCSVV2 >= B_TAG_THRESHOLD, axis=1) == 1)
selections.add("atleast_2b", ak.sum(jets.btagCSVV2 > B_TAG_THRESHOLD, axis=1) >= 2)

  • 👍1
eguiraud commented

Pointed out by @andriiknu actually 😬

According to the spec it should be > everywhere: https://agc.readthedocs.io/en/latest/versionsdescription.html

  • 👍1
alexander-held commented

#174 is dedicated to tracking the b-tagging.

alexander-held commented

While doing this (which will trigger the need for new reference histograms), also remove the offsetting by 1e-6 in

analysis-grand-challenge/analyses/cms-open-data-ttbar/utils/file_output.py

Lines 8 to 9 in ca06789

for channel in channel_names:
hist_dict[channel] += 1e-6
(this was a workaround to avoid issues due to empty bins for the statistical inference stage).