Efficiency and fake rate of the cut-based electron ID for Run 2.

Samples

Signal samples from the CSA production can be found in DAS, by typing in the query window:

• dataset = /RSGravToZZ*/*Spring14dr*/AODSIM

To implement the cut-based electron identification we produced private MINIAODSIM samples with few events (100K), using the common ID and Isolation Toolkit CITK.

To access the private samples from lxplus, you have to obtain a valid grid proxy (voms-proxy-init --voms cms). The process.source in your configure file should look like the following:

process.source = cms.Source("PoolSource",
    fileNames = cms.untracked.vstring('root://cmsxrootd.fnal.gov//store/user/jruizvar/RSGravToZZ/M1000/RSGravToZZ_kMpl01_M-1000_Tune4C_13TeV-pythia8_MINIAODSIM_PU20bx25_1.root')
)

Matching categories

MINIAODSIM samples contain information about generated and reconstructed particles. Our matching algorithm identifies the closest genElectron for every recoElectron, as well as the particle ID of the genElectron's ancestor.

The deltaR and PID matching allows to define four categories as described in the following table:

Efficiency and fake rate

The efficiency is the number of signal electrons passing the ID divided by the total. The fake rate was estimated in a similar way but using the non-signal electrons.

To access the electronID, which is a member function of the pat::Electron class, we used a line like the following:

int passID = ( el->electronID("cutBasedElectronID-CSA14-PU20bx25-V0-standalone-veto") > 0.5 );

Working points

The cut-based electron identification recommended by EGM for Run 2 analyses, determines four working points: veto, loose, medium, and tight. A detailed description here.

Isolation

Particle flow isolation with delta beta correction is calculated as follows:

    reco::GsfElectron::PflowIsolationVariables pfIso = el->pfIsolationVariables();
    absiso = pfIso.sumChargedHadronPt + std::max(0.0 , pfIso.sumNeutralHadronEt + pfIso.sumPhotonEt - 0.5 * pfIso.sumPUPt );
    relIsoWithDBeta = absiso/pt;

Internally, the isolation variables use the standard recommended cone size deltaR = 0.3. Details here.

* Electron isolation after different cut-based selections

Preliminary results

To estimate the efficiency and fake rate of the electron identification, we classified the electrons according to its matching category and the passID associated to the cutBasedElectronID-CSA14 working points.

The classification for each working point is displayed as a TH2 histogram with two bins in the Y axis associated to the passID variable, and four bins in the X axis corresponding to the matching categories. The number of entries inside each box of the histogram indicates the number of electrons in a given matching category with a given passID.

• Electron classification, global efficiency, and fake rate using sample RSGravToZZ 1 TeV
  

• Electron classification, global efficiency, and fake rate using sample RSGravToZZ 4.5 TeV
  

Efficiency and fake rate as function of genElectron's pT and eta are shown below:

• Efficiency and fake rate as function of genElectron's pT
  

• Efficiency and fake rate as function of genElectron's eta
  

• Efficiency as function of deltaR(genElectron1, genElectron2)
  

• Efficiency as function of deltaR(genElectron1, genElectron2) zoom in the low deltaR region
  

Software

The analyzer developed for these studies is attached below:

• Configure file
• EDAnalyzer
• BuildFile

Ongoing work

In addition to the cut-based electron ID, we plan to use the Multivariate Electron Identification for Run 2. Instruction for running the MVA electron ID in the following link.

Comments