Skip to content

Bad muon scale factors in 2023/2024

Opening this issue is to investigate wrong scale factors/muons being applied in 23/24 as spotted by the CAT CI.

For 2024:

  • Mini: /eos/cms/store/group/cat/datasets/recipes/24CDEReprocessingFGHIPrompt-Summer24/TTTo2L2Nu_10k_mini.root
  • Nano: /eos/cms/store/group/cat/datasets/recipes/24CDEReprocessingFGHIPrompt-Summer24/TTTo2L2Nu_10k_nano.root

Standard ntuplisation followed by mergeNtuples and the muon scale factors command:

$ applyMuonSelection -f M1_merge.root M2_muon.root /cvmfs/cms-griddata.cern.ch/cat/metadata/MUO/Run3-24CDEReprocessingFGHIPrompt-Summer24-NanoAODv15/latest/muon_Z.json.gz:NUM_TightID_DEN_TrackerMuons

CAT complains:

2690 event(s) found with wrong output values, first five:
  (188516294, 1, 253043): truth=(0.9620501,), result=(0.952762,)
  (188516308, 1, 253043): truth=(0.9920142,), result=(0.987405,)
  (188516300, 1, 253043): truth=(0.9826044,), result=(0.983185,)
  (188516314, 1, 253043): truth=(0.98339486,), result=(0.980044,)
  (188516305, 1, 253043): truth=(0.9735898,), result=(0.963302,)

Other frameworks using the exact same input file get consistent results, suspecting a difference between the ntuple and the NanoAOD.

  • Tried changing the global tag to match NanoAOD, same issue.
  • Comparing muons for event 188516294, they have identical kinematics in ntuple and Nano: 
    root [11] events->Scan("evtNo:recMuons.p4.Pt():recMuons.p4.Eta():recMuons.Weight()")
***********************************************************************
*    Row   * Instance *     evtNo * recMuons. * recMuons. * recMuons. *
***********************************************************************
...
*        1 *        0 * 188516294 * 51.383045 * 0.7878287 * 0.9915292 *
*        1 *        1 * 188516294 * 16.203302 * -1.059794 * 0.9609014 *
  • I find that truth/recMuons[0].Weight() = 0.9620501/0.9915292 = 0.9702689 is present in the JSON file.
  • Suspecting eta vs |eta|: the 2024 scale factors are binned in signed eta...