Higgs to Zg to llg analysis with run2 data

Questions raised in July 22

- L1 prefired issue for 2016/2017

We followed the instruction from JETMET group to assign a weight for the events suffer from L1 prefired problems

- VBF jets purity of signal sample

We use the LHE information to identify the true VBF jets, the purity of the VBF jets is 65%. The VBF jets preselections considers the events with highest jet pT. The matched VBF jets input to MVA training as the signal. This gives significant improvement on the ROC curve.

- MVA uncertainties

consider the uncertainties from the training variables propagating to the shape of MVA. We fixed the MVA score cuts , the difference of signal yield by the variation of the MVA shape is considered as the uncertainty

kinematic MVA-

  • R9 reweighting:Since this shower-shape variable does not match in data and MC, we reweight the MC to the r9 distribution in data. With this there comes an uncertainty associated to the re-weighting. To estimate it, we remove the r9 re-weighting and re-estimate the change in yields in each of the four untagged categories (since r9 variables is the input of kinematic MVA which is used to do the categorization for untagged events).
  • Photon MVA score: The prescription which is followed to assign a systematic uncertainty to this variable is to shift its value for every photon in the simulation according to a transformation combining a shift of 0.03 with a linear increase of the uncertainty for events with a low score from the photon ID MVA.Since the value of the kinematic MVA tends to be larger for larger values of the photon ID MVA,the simultaneous translation of the photon ID MVA for both photons leads to the maximal migration of events in the diphoton MVA output, which is then propagated as a migration of the signal yield among the final event categories.
  • Photon resolution:The per-photon resolution estimate is also affected by imperfect modelling of the electromagnetic shower shape in the simulation. This quantity affects the kinematic MVA output. The impact is diluted since it competes in the sum in quadrature with the simulated energy resolution smearing term. The systematic uncertainty from modelling of σE/E is assigned shifting its value by 5% for each photon.
  • lepton momentum scale correction: Due to the pT measurements affect pTllγ/mllγ measurement, we consider the electron&muon scale uncertainty to get the new MVA distribution
  • Jet energy correction:Jet energy corrections are applied to data and simulation as appropriate. The uncertain- ties on the corrections are applied to simulation and used to derive the magnitudes of shifts between VBFTag0 and VBFTag1, as while as between the VBF tags and the untagged events. Nusiance parameters are defined such that migrations between the two VBF tags can be adjusted independently from those to untagged events in the final fit.
  • Jet energy resolution:Jet energy resolution corrections are applied to simulation by using recommended scale factors, which are applied to $pT_{reco} − pT_{true}$. In the case of no truth-level jet being matched, an additional smearing is applied in accordance with the JetMET group recommendations. The results of these shifts are implemented as migrations between bins similarly to the jet energy corrections.
- Uncertainty correlation for different years:
  • Theoretical uncertainties and parton showering uncertainty: 100% correlated between 3 years.
  • Underlying events: 100% correlate between 2017 and 2018 MC, and uncorrelate with 2016.
  • Luminosity: follow the recommendation of lumiPOG, considered the correlated/uncorrelated parts.
  • Pileup reweighting: Assuming 100% correlation due to the theoretical assumption is correlated.
  • Trigger efficiency: The uncertainties are negligible that we assume the correlation should be 0%
  • ID efficiency Considered 100% correlation from EGM recommendation
  • Conversion safe electron veto:0% correlation due to the different pixel detector / pixel running conditions between the years.
  • EGM energy scale: The statistical and gain components are uncorrelated, the systematic part is 100% correlated.
  • EGM energy resolution: The errors are 100% correlated between the years as they are a measurement and the same method is used in all years.
  • Rochester muon momentum correction: Statistical part of both scale and resolution are uncorrelated. The rest components are 100% due to the methods are the same.
  • R9 reweighting: Assume 100% correlated due to the methods are the same.
  • Photon ID MVA: Assume 100% correlated due to the methods are the same.
  • Photon resolution: Assume 100% correlated due to the methods are the same.
  • Jet energy scale and resolution: The recommendation is part correlate between 3 years in jet energy scale, for simplicity we assume 100%

Questions raised in November 25

  • Need to document the shower shape correction in AN include validation.

  • need to get the nuisance impact plot.
We need to update this after we have all uncertainties. -- MingYanLee - 2019-09-05
Edit | Attach | Watch | Print version | History: r5 < r4 < r3 < r2 < r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r5 - 2019-12-09 - MingYanLee
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Sandbox All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2023 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
or Ideas, requests, problems regarding TWiki? use Discourse or Send feedback