Reduced event output (Ntuples or TTrees or EDM events?), what is to store there?

[xealits]

Can output an NTuple of particular parameters for preselection or full EDM event for a particular selection.

NTuple has some limitations, so in principle a TTree of object would be better.

A simple TTree restricted to a list of float variables only. Each variable goes to a separate branch. A Ntuple is created via

TNtuple(name,title,varlist,bufsize)

It is filled via:

TNtuple::Fill(*x)  or
TNtuple::Fill(v1,v2,v3.....)

and

Create an Ntuple. The parameter varlist describes the list of the ntuple variables separated by a colon: Example: x:y:z:energy For each variable in the list a separate branch is created. NOTE:

• Use TTree to create branches with variables of different data types.
• Use TTree when the number of branches is large (> 100).

[xealits]

What to put in the NTuple:

• the current control plots and selections:
  • leading leptons, jets, b-jet
    • pdgId, eta, phi, pt, p
    • for tau: IDlev (highest tau ID: 1 - Loose, 2 - Medium, 3 - Tight)
    • for jets: radius [split into etaetaMoment and phiphiMoment], bDiscr, hadronFlavour (b-tagging), paronFlavour (fakes) {could add Jet PU discriminators}
  • met: uncorrected/corrected + control over 2016 issues
  • N particles in the event (N lept, N jets, N bjets, N taus)
  • different rho, N vertices, true N vert in MC (0 in Data)
  • aMC@NLO -1 weight
  • Top pT weight
• add "channel ID" = product of lepton IDs (just lepton ID if single-lepton)
• Top channel (from generation info -> different output files, as is now)
• Nup for WNJets
• for anti-lj:
  • if output all initial jets, then their b-discriminant is needed
  • momentum of tau & tau-jet
  • tau+light-jet mass closest to W peak & tau+light-jet+b-jet mass closest to t peak (output these two as a test; previous -- can calculate from NTuple)
• for fake rates:
  • jet origins (partonFlavour & match to gen taus)
• for fake factor:
  • number of loose/medium/tight taus ("level" of ID -- set 1/2/3 depending on maximum ID, loose/medium/tight)
  • their matches to jets? (no, for now)
  • can a fake factor approach be done to b-tagging? loose/medium/tight b-tag? -> output b-tag discr
• vertices
  • N good vertices (maybe parameters of "goodness"?)
  • covariance matrix
• for tau ID:
  • R variable of the jet (leadin track energy / all energy)
  • tau decay channel (N charged hadrons)
  • info on constituent tracks, impact parameters, secondary vertex
  • there are internal variables of tau ID algorithm available -- some of them?
• for luminosity: run and lumisection -> dump in json-s for now, figure out later
• MET > 20 cut & >=1 jet to reduce the output a bit
• add mass of dilepton system > 10-20 GeV?
• add electron/muon rel iso for background regions?

-- could output certain N of jets and counters for the rest (like 5 selJetsNoLep jets are in output, with their momenta, b-tag discriminants etc and also output N of all jets, ID jets, selJets, selJetsNoLep jets etc). And make 2 different NTuples: for single-lepton and dileptons.