Run3 HH4b Cutflow Challenge

• on cmslpc335

• Remaking B from nano
  • python runner.py -s -p skimmer/processor/skimmer_4b.py -c skimmer/metadata/HH4b.yml -y 2022_preEE -d data -e B -op skimmer/metadata/ -o picoaod_datasets_data_2022_preEEB_debug_v2.yml -m metadata/datasets_HH4b_Run3.yml --condor
  • python3 metadata/merge_yaml_datasets.py -m metadata/datasets_HH4b_Run3.yml -o metadata/datasets_HH4b_Run3_merged_debug_v2.yml -f skimmer/metadata/picoaod_datasets_data_2022_preEEB_debug_v2.yml
  • python runner.py -s -p skimmer/processor/skimmer_4b.py -c skimmer/metadata/HH4b_fourTag.yml -y 2022_preEE -d data -op skimmer/metadata/ -o picoaod_datasets_fourTag_data_Run3_vDebug_v2.yml -m metadata/datasets_HH4b_Run3_merged_debug_v2.yml --condor
• Events not in four b skimmmer… problem was eta requirement in lowPt skimming
• Fixed.
• Now re-running skims on C and D over all
• Re-doing the four-tag skim
• Agreement now:
  • Total number of missing events: 317 (1.3% of total events)
  • Total number of extra events: 5986 (19.87% of total events)
• Going to reskim all Run3
  • Submitted 2022_EE : source skim2022_EE.sh

• I think the cutflow problem could be due to the four tag skimmer.
• merging:
  • python3 metadata/merge_yaml_datasets.py -m metadata/datasets_HH4b_Run3.yml -o metadata/datasets_HH4b_Run3_merged_debug.yml -f skimmer/metadata/picoaod_datasets_data_2022_preEE*_debug.yml
• Running with out 4b skim on cmslpc331
  • python runner.py -o data_Run3_vDebug_new.coffea -d data -p analysis/processors/processor_HH4b.py -y 2022_preEE -op output/synthetic_dataset_analyze_all -c analysis/metadata/HH4b_run_fastTopReco.yml -m metadata/datasets_HH4b_Run3_merged_debug.yml --condor
• taking too long…. Will do 4b skim
  • python runner.py -s -p skimmer/processor/skimmer_4b.py -c skimmer/metadata/HH4b_fourTag.yml -y 2022_preEE -d data -op skimmer/metadata/ -o picoaod_datasets_fourTag_data_Run3_vDebug.yml -m metadata/datasets_HH4b_Run3_merged_debug.yml --condor
• Merging
  • python metadata/merge_yaml_datasets.py -m metadata/datasets_HH4b_Run3.yml -o metadata/datasets_HH4b_Run3_fourTag_vDebug.yml -f skimmer/metadata/picoaod_datasets_fourTag_data_Run3_vDebug.yml
• Running
  • =python runner.py -o data_{Run3vDebugnew.coffea} -d data -p analysis/processors/processor_HH4b.py -y 2022_preEE -op output/synthetic_{datasetanalyzeall} -c analysis/metadata/HH4b_{runfastTopReco.yml} -m metadata/datasets_{HH4bRun3fourTagvDebug.yml} =
• See way too few events in debug… needed to update the debug mask… Rerunning
• Agreement now
  • Total number of missing events: 1119 (4.58% of total events)
  • Total number of extra events: 5778 (19.84% of total events)
• Most events dont make the skim…
• Reskimming on (cmslpc324) with debug info for correct numbers:
  • python runner.py -s -p skimmer/processor/skimmer_4b.py -c skimmer/metadata/HH4b_fourTag.yml -y 2022_preEE -d data -op skimmer/metadata/ -o picoaod_datasets_fourTag_data_Run3_vDebug.yml -m metadata/datasets_HH4b_Run3_merged_debug.yml --condor
• Jet eta cut is different! 2.4 (US) vs 2.5 (Marina)
• Reskimming with 2.5 cut
  • Total number of missing events: 580 (2.37% of total events)
• Added abs(delta_dhh)
• Remaking the skims with the new eta cut
  • Ran B, C and D
  • python runner.py -s -p skimmer/processor/skimmer_4b.py -c skimmer/metadata/HH4b.yml -y 2022_preEE -d data -e B -op skimmer/metadata/ -o picoaod_datasets_data_2022_preEE_debug.yml -m metadata/datasets_HH4b_Run3.yml --condor
• Merged .. Remade 4b… (Not clear where the events are going…they appear as fourTag in the first skim)
• Hmmm back to 5% …

• cmslpc319
• Skim works when running on condor !
• All events are there in the skim now!
• Will re run C and D before redoing the cutflow
• python runner.py -s -p skimmer/processor/skimmer_4b.py -c skimmer/metadata/HH4b.yml -y 2022_preEE -d data -e C  -op skimmer/metadata/ -o picoaod_datasets_data_2022_preEEC_debug.yml -m metadata/datasets_HH4b_Run3.yml --condor

Run-3 4b skims

• checking which file has the run I care about
• dasgoclient --query"run dataset=/JetMET/Run2022C-22Sep2023-v1/NANOAOD"=
• > python runner.py -s -p skimmer/processor/skimmer_4b.py -c skimmer/metadata/HH4b.yml -y 2022_preEE -d data -e B -op skimmer/metadata/ -o picoaod_datasets_data_2022_preEE_debug.yml -m metadata/datasets_HH4b_Run3.yml --condor
• Running on cmslpc332
• Adding debugging output

• Impletented the Run3 case when dhh < 30 GeV
• Remade pkl inputs
• Now better agreement!
  • Before: Total number of missing events: 11690 (47.79% of total events) Total number of extra events: 4918 (27.81% of total events)
  • Now: Total number of missing events: 8426 (34.45% of total events) Total number of extra events: 2706 (14.44% of total events)
• Eg: Run 355872: 99 missing events =First 5 missing events: [228793492, 993131619, 994362241, 994277091, 1003079747, 1007831198, 1141642728, 1145575379, 1145596930, 1086081738]

• Yes, we follow the method used in HIG-20-005 ( [https://gitlab.cern.ch/mkolosov/hh4b_run3/-/blob/run2/python/producers/hh4bTreeProducer.py?ref_type=heads#L3350](https://gitlab.cern.ch/mkolosov/hh4b_run3/-/blob/run2/python/producers/hh4bTreeProducer.py?ref_type=heads#L3350) )  in which :  if the difference of the minimum and second minimum distance from the diagonal is less than 30, we choose the pair that has the maximum H1 pT in the 4-jet center-of-mass frame.
• Implementing mhh < 30 check
• python -m unittest analysis.tests.quadjet_test.quadJetTestCase

• finding events…
• Most (70%) missing the skimmer.
• lxplus down…not down just slow
• Finding problem events… printing out the jet four vectors

• adding event and run to the skim
• 50% of events different !
• run 355872: 94 missing events First 10 missing events: =[228793492, 993131619, 994277091, 995618735, 1003079747, 1145575379, 917646697, 1086081738, 276519899, 321185192 …
  • python runner.py -o synthetic_data_Run3_v8_debug_seedXXX.coffea -d data -p analysis/processors/processor_HH4b.py -y 2022_preEE -op output/synthetic_dataset_analyze_all -c analysis/metadata/HH4b_run_fastTopReco.yml -m metadata/datasets_HH4b_Run3_fourTag_v8.yml
• Errors with running in cofffea … debugging (in Cursor! )

• made python script to read Marinas file (Cursor !!!)
  • = /Users/johnda/CMSWork/hh4b/coffea4bees/python/analysis > py process_{eventnumbers.py} =
• re-Making my pkl file:
  • python runner.py -o synthetic_data_Run3_v8_debug_seedXXX.coffea -d data synthetic_data -p analysis/processors/processor_HH4b.py -y 2022_preEE -op output/synthetic_dataset_analyze_all -c analysis/metadata/HH4b_run_fastTopReco.yml -m metadata/datasets_HH4b_Run3_fourTag_v8.yml
• Marina's file has all the data events… How to get only the 4b CR events ? pinged her
• Got instructions for CR events
• skipping the input file for 4Tag events
• fighting with trying to filter the 4b events

• Marina's cutflow is off
  • requested events/run list
  • [>>] Compare Events

• added pt_raw to debug output
• remade pkl file for Marina… sent

• Fixed Cutflow for Marina
• Add was only including one data Era
• Add other eras
• Add other Pt corrections
• sent

• Implementing Run3 Selection
• Working cutflow … got unit test working
• Adding dhh min
• Synch Higgs Candidate Selection
• Making Cut flow for Marina:
  • python runner.py -o synthetic_data_Run3_v8_debug_seedXXX.coffea -d data synthetic_data -p analysis/processors/processor_HH4b.py -y 2022_preEE -op output/synthetic_dataset_analyze_all -c analysis/metadata/HH4b_run_fastTopReco.yml -m metadata/datasets_HH4b_Run3_fourTag_v8.yml
• Make output files for Marina
  • python analysis/debugRun3SyntheticData.py output/synthetic_dataset_analyze_all/synthetic_data_Run3_v8_debug_seedXXX.coffea
• Sent cutflow to Marina
• Running plots for data vs synthetic

• Implementing Run3 selection
• From Marina:
  • Code: https://gitlab.cern.ch/mkolosov/hh4b_run3/-/blob/run2/python/producers/hh4bTreeProducer.py?ref_type=heads
  • Config: https://gitlab.cern.ch/mkolosov/hh4b_run3/-/blob/run2/run/configs/2021/main/run3_2021_0L_PNet_MinDiag_Jets_4j35_2bj30_dHHjw30.yaml?ref_type=heads
• Synch Lepton Definition
• [>>] Synch Higgs Candidate Selection
• Synched Electrons
• Synched Muons:
• Testing with
  • time python runner.py -t -o synthetic_data_Run3_2022_preEE.coffea -d data -p analysis/processors/processor_HH4b.py -y 2022_preEE -op output/synthetic_dataset_analyze_all -c analysis/metadata/HH4b_run_fastTopReco.yml -m metadata/datasets_HH4b_Run3_fourTag_v6.yml
• Trying test case
  • python -m unittest analysis.tests.quadjet_test.quadJetTestCase

• output event list for florida
  • python debug_HHRun3.py
• Outputting our preSelection event list
  • time python runner.py -o data_Run3_2022_EE.coffea -d data -p analysis/processors/processor_HH4b.py -y 2022_EE -op hists -c analysis/metadata/HH4b_run_fastTopReco.yml -m metadata/datasets_HH4b_Run3_fourTag_v4.yml
• Ask Marina to compare unblinded inclusive 4b counts… Maybe normalization is CR acceptance difference
• Made our event numbers list
• python analysis/debugRun3CutFlow.py hists/data_Run3_2022_EE.coffea
• Only missing ~1% of events
• Some differences in jet and b-jet multiplicities

• Copied File from marina to root://cmseos.fnal.gov//store/user/jda102/debugRun3HH/data_2022_postEE_0_tree.root
• Chat GPT to write uproot program
• Wrote out a pkl file with the events and lumis
• Make pickle of events passing 4jets selection