4.2.1 Physics Analysis Toolkit (PAT): Data Formats

Data type to describe reconstructed photons, which is derived from the reco::Photon. In addition to the reco::Photon it may contain the following extra information:

• isolation calculated in a pre-defined cone in different sub-detectors of CMS. You can choose between isolation as recommended by the corresponding POG or a user defined isolation accessible via the member function userIsolation(pat::IsolatioKey key), where the available IsolationKeys are defined in Isolation.h. Note: user defined isolation needs to be configured during the pat::Candidate creation step as described in WorkBookPATConfiguration.
• isolation deposits for user defined isolation and more detailed studies
• match on generator level
• match on trigger level
• reconstruction efficiency
• object resolution
• Id variables
• shower shape variables

For a description of the algorithms to determine this extra information have a look at WorkBookPATWorkflow.

The labels of all photon collections, which are created during the standard PAT workflow(s) are given below:

Data Type	Collection Label	Instance Label	Process Label
pat::PhotonCollection	patPhotons	None	PAT
pat::PhotonCollection	selectedPatPhotons	None	PAT
pat::PhotonCollection	cleanPatPhotons	None	PAT

For a description of each listed module have a look at WorkBookPATConfiguration.

Data type to describe reconstructed electrons, which is derived from the reco::Electron. In addition to the reco::Electron it may contain the following extra information:

• isolation in a predefined cone
• isolation deposits for user defined isolation and more detailed studies
• match on generator level
• match on trigger level
• reconstruction efficiency
• object resolution
• Id variables
• shower shape variables

The electrons are sorted in descending order in p_t. For a description of the algorithms to determine this extra information have a look at WorkBookPATWorkflow.

The labels of all electron collections, which are created during the standard PAT workflow(s) are given below:

Data Type	Collection Label	Instance Label	Process Label
pat::ElectronCollection	patElectrons	None	PAT
pat::ElectronCollection	selectedPatElectrons	None	PAT
pat::ElectronCollection	cleanPatElectrons	None	PAT

For a description of each listed module have a look at WorkBookPATConfiguration.

Data type to describe reconstructed muons, which is derived from the reco::Muon. In addition to the reco::Muon it may contain the following extra information:

• tracks from a TeV refit (pat::Muon::pickyMuon(), pat::Muon::tpfmsMuon())
• isolation in a predefined cone
• isolation deposits for user defined isolation and more detailed studies
• match on generator level
• match on trigger level
• reconstruction efficiency
• object resolution
• two MuonMETCorrectionData objects. These objects allow to correct/uncorrect the caloMET and tcMET on the fly for the pat::Muon. They are obtained from the Muon-MET correction value maps that are described on WorkBookMetAnalysis#HeadingThree and WorkBookMetAnalysis#HeadingFour.

The muons are sorted in descending order in p_t. For a description of the algorithms to determine this extra information have a look at WorkBookPATWorkflow.

The labels of all muon collections, which are created during the standard PAT workflow(s) are given below:

Data Type	Collection Label	Instance Label	Process Label
pat::MuonCollection	patMuons	None	PAT
pat::MuonCollection	selectedPatMuons	None	PAT
pat::MuonCollection	cleanPatMuons	None	PAT

For a description of each listed module have a look at WorkBookPATConfiguration.

Data type to describe reconstructed taus, which is derived from the reco::Tau. In addition to the reco::Tau it may contain the following extra information:

• isolation in a predefined cone
• match on generator level
• match on trigger level
• reconstruction efficiency
• object resolution
• tau discrimination variables

The taus are sorted in descending order in p_t. Per default particle flow taus are used, though this may be switched to calorimeter taus during the step of pat::Tuple creation. For a description of the algorithms to determine this extra information have a look at WorkBookPATWorkflow.

The labels of all tau collections, which are created during the standard PAT workflow(s) are given below:

Data Type	Collection Label	Instance Label	Process Label
pat::TauCollection	patTaus	None	PAT
pat::TauCollection	selectedPatTaus	None	PAT
pat::TauCollection	cleanPatTaus	None	PAT

For a description of each listed module have a look at WorkBookPATConfiguration.

Data type to describe reconstructed jets, which is derived from the reco::Jet. In addition to the reco::Jet it may contain the following extra information:

• b tag information
• jet energy corrections
• jet charge
• associated tracks
• jet flavour
• match on generator level (reco::genJet und parton)
• match on trigger level
• reconstruction efficiency
• object resolution

Per default anti-kT jets are used with R=0.5, though this may be switched to any other kind of jet collection during the step of pat::Tuple creation. Also alternative jet collections may be added to the standard pat::Tuple content depending on the analysis purposes. The energy scale of the pat::Jets is corrected to L3Absolute, while the correction factors for L2Relative, L5Flavour and L7Parton are also stored within the jet. They may be used to produce a clone of the jet, which is corrected to the specified level or fully uncorrected ('raw'). For more details on the arguments to be used have a look at the description of the pat::JetCorrFactors below. The jets are sorted in descending order in (L3Absolute calibrated) p_t. For a description of the algorithms to determine this extra information have a look at WorkBookPATWorkflow.

The labels of all jet collections, which are created during the standard PAT workflow(s) are given below:

Data Type	Collection Label (1)	Instance Label	Process Label
pat::JetCollection	patJets	None	PAT
pat::JetCollection	selectedPatJets	None	PAT
pat::JetCollection	cleanPatJets	None	PAT

(1) Note that when adding jet collections to the event content of the standard pat::Tuple the additional jet collections will be accompanied by a postfix identifying the additional jet collection. This postfix typically could be of type cleanPatJetsAK5Calo.

For a description of each listed module have a look at WorkBookPATConfiguration.

Data type to describe reconstructed MET, which is derived from the reco::MET. In addition to the reco::MET it may contain the following extra information:

• type 1 MET corrections
• muon MET corrections
• match on generator level
• match on trigger level
• reconstruction efficiency
• object resolution

Where the type 1 MET correction corresponds to the standard jet selection in use. Per default pat::MET is type 1 corrected and muon corrected. The muon corrections may be altered on the fly depending on the selection criteria for muons. For more details have a look at the description of the pat::Muon. For a description of the algorithms to determine this extra information have a look at WorkBookPATWorkflow.

The labels of the possible MET collections produced in a PAT workflow are given below:

Data Type	Collection Label	Instance Label	Process Label
pat::METCollection	patMETs	None	PAT
pat::METCollection	patMETsTC	None	PAT
pat::METCollection	patMETsPF	None	PAT

The patMETs collection is produced in the standard workflow, but the track corrected MET collection (patMETsTC) and the particle flow MET collection (labeled as patMETsPF) may be added to the content of the standard pat::Tuple on the user's choice (see patTuple_addJets_cfg.py ). For a description of each listed module have a look at WorkBookPATConfiguration.

Data type to host all potential correction factors of the factorised calibration ansatz (L1 to L7) to be embedded in the pat::Jet. More then one set of jetCorrFactors can be embedded in one pat::Jet, as useful for the comparison of different sets of calibrations or systematic studies. The pat::JetCorrFactors class will be issued internally when creating clones of pat::Jets with different calibration levels. This will need the correction step and the potential flavour as arguments. The following correction levels (and flavours) are supported:

correctstep	flavour	correction type
Uncorrected	-	uncalibrated jet
L1Offset	-	offset correction
L2Relative	-	relative inter eta correction
L3Absolute	-	absolute pt correction
L23Residual	-	residual corrections between data and MC after the L3Absolute correction
L4Emf	-	correction as a function of the jet emf
L5Flavor	gluon uds charm bottom	hadron level correction for gluons, light quarks, charm, bottom
L6UE	gluon uds charm bottom	underlying event correction for gluons, light quarks, charm, bottom
L7Parton	glu uds charm bottom	parton level correction for gluons, light quarks, charm, bottom

Note: the flavour strings are case insensitive. Per default the set of jetCorrFactors is embedded in the pat::Jet, which corresponds to the JetMET default and the corrections for L1Offset, L2Relative, L3Absolute, L5Flavor and L7Parton are stored in addition. All other correction factors will be transparent. Per default the pat::Jet will be corrected to the following level (if available from the list of correction levels) with the following priorities:

• to L2L3Residual if available in the list of correction levels.
• to L3Absolute if available in the list of correction levels.
• to Uncorrected (i.e. no correction will be applied, if niether L2L3Residual nor L3Absolute are available in the list of corrction levels.

In the latter case, the PatJetProducer will issue a Warning during patTuple production. For a description of the algorithms to determine this extra information have a look at WorkBookPATWorkflow.

The labels of all jet collections, which are created during the standard PAT workflow(s) are given below:

Data Type	Collection Label	Instance Label	Process Label
pat::JetCorrFactors	patJetCorrFactors	None	PAT

For a description of the module have a look at WorkBookPATConfiguration.

Data type to host potential object resolutions. It can be embedded to each pat::Object. For a detailed description have a look at SWGuidePATResolutions.

Note that adding object resolutions and embedding them into a pat::Candidiate is OPTIONAL. They are not part of the default PAT event content. To add resolutions to an arbitary pat::Candidate add a service to your cfg file for patTuple production and switch the embedding of the resolution information for the corresponding pat::Candidate to True. An example for b jets (in one central bin in eta) is given below:

## define the service
bjetResolution = stringResolution.clone(parametrization = 'EtEtaPhi',
    functions = cms.VPSet(
      cms.PSet(
        bin = cms.string('0.000<=abs(eta) && abs(eta)<0.087'),
        et  = cms.string('et * (sqrt(0.0901^2 + (1.035/sqrt(et))^2 + (6.2/et)^2))'),
        eta = cms.string('sqrt(0.00516^2 + (1.683/et)^2)'),
        phi = cms.string('sqrt(0.0024^2 + (3.159/et)^2)'),
      ),
      # ... in reality you will have more bins in eta here of course ...
    ),
    constraints = cms.vdouble(0)
)

# embed the resolutions inot the patJets collection
patJets.addResolutions = True
patJets.resolutions    = bjetResolution

You can find an example cff file containing resolutions for muons, jets and MET as recently derived within the TopPAG based on MC truth information in a typical madgraph ttbar sample in stringResolutions_etEtaPhi_cff.py in the python/recoLayer0 directory. Please note that this is ONLY AN EXAMPLE file to demonstrate the technique. It does not ease you from the burden to derive and test the resolutions which might be applicable and adequte for your analysis.