Open ywan2 opened 1 year ago
TODO (based on Kelci's comments on November 10 meeting): Add a flag to be able to switch on/off these new categorizations instead of completely wiping off the baseline categorizations employed in Top-22-006.
I had some free time, so I fixed the lint
error, and merged in the latest master branch updates. I think we just need the flag to turn these changes on/off for validation, then we should be good to merge.
TODO (based on Kelci's comments on November 10 meeting): Add a flag to be able to switch on/off these new categorizations instead of completely wiping off the baseline categorizations employed in Top-22-006.
There are a coupe of ways we could do this flag. The simplest might actually be in the datacard maker, where we just sum together the 3l off-Z
low
and high
bins. @ywan2 you could work with @Andrew42 to figure this out. Otherwise, we could just put logic in the processor to use the old or new binning.
@ywan2 is get_off_Z_mask_low
in your topcoffea
repo, or did you move it over to topeft
?
https://github.com/TopEFT/topeft/actions/runs/6826768955/job/18567397360?pr=385#step:12:161
In order to improve the sensitivity of our analysis (through diminishing 1D likelihood 2sigma confidence intervals), we want to divide up some event categories - to split
3l_offZ*
channels into3l_offZ_high*
and3l_offZ_low*
. This PR is created with the following main changes:In
analysis_processor.py
:sfosz_3l_OffZ_low_mask
modules/event_selection.py
script intopcoffea
package.selections.add()
now has double the number of offZ categories3l_x_offZ
categories to 4offZ_low
and 4offZ_high
categories.offZ_low
has extra mask (mentioned above) to put everything outside on_Z in the low mass region.offZ_high
includes the anti-mask to put what is left in the high mass region.sr_cat_dict
within3l
,3l_offZ*
channels are changed into3l_offZ_high*
and3l_offZ_low*
In
datacards_post_processing.py
:README is also updated with necessary instructions.
It is important to know that this PR is intertwined with topcoffeaPR#10, so both have to be implemented for the analysis to run.