Tau Efficiency Measurement using Tag and Probe with Ztautau events.


Page for Tau Eff stuff

Public Results:



First Pass: (By mid June)

  • Quick re-optimisation if necessary on 2011 data
  • Extract single efficiency number with ntrack and |charge| in tau selection to supress QCD events
  • Run on Period A-D (160pb-1 with no LAr problem)
  • Evaluate Systematics
  • Finalise first Tau Eff measurement with channel

Second Pass: (late summer)

  • Include Electron Leg
  • Include all available data (including LAr problem)
  • Extract 1P and 3P/MP efficiency
  • Release ntrack and |charge| requirement from BaseTau selection for inclusive measurement.
  • Include Lepton Isolation fit for QCD estimation


  • Tau Trigger measurement:
    • Use Analysis including Tau ID to get pure true tau sample
    • Read 2010 note here
  • Tau Scale measurement:
    • Also use tight tau id to get very pure tau sample, then fit vis mass dist.

Useful Links

Z->tautau links

Tau Links:

ATLAS Links:

Analysis Code

Analysis flow:
  • Run on tauPerf D3PDs
  • Process with ZToTwoTausAnalysis (modified for Tau Eff measurement) - NOTE: Should we chuck our modified code in a repository somewhere?
  • Dump out summary TTrees with derived quantities for data and MC samples
  • Process TTrees with TauEffPyTools - the package we developed for the analysis.

Links to Code:

2011 Data Info

  • GRLS (NEW):
    • Extracted GRLs for specific periods from Official GRL 'data11_7TeV.periodAllYear_DetStatus-v16-pro08-03_Z_tautau_lh.xml', using 'dqu_grl_per_period.py'.
      • Output GRLs:
        • [[][PeriodA]]
        • [[][PeriodB]]
        • [[][PeriodD]]
        • [[][PeriodE]]
    • Ran Lumi Calc on GRLs:
      • Settings:
      • Results:
        • Period A, no events, L = 0 pb-1
        • Period B, [[][result]], [[][histos]], L = pb-1
        • Period D, [[][result]], [[]], L = 147.963 pb-1
        • Period E, [[][result]], [[][histos]], L = 41.6104 pb-1
        • Period F, [[][result]], [[][histos]], L = 124.49 pb-1

2011 Recommendations

2011 Code


High Priority

  • Check kW with OS/SS events

  • Check Vertex Counting!
  • Check pileup reweighting

  • Muon isolation measurement using Zmumu Tag & Probe (David):
    • Read Appendix B in Ztautau cross-section support note
    • Redo analysis using our selection for probe.
    • If we are going to do isolation fitting, we might need to look at actually re-weighting MC iso dist, rather than just applying efficiency corrections. But that is second iteration.

  • Improving QCD estimates:
    • Fitting Isolation distributions:
      • Use e-mu QCD control region to get muon isolation shape for QCD events.

    • Using 2D fit with 2 good QCD discriminators to extract QCD contribution
    • Check Ros/ss correlations vs Isolation
    • Fix TauEffCycle:
      • remove ntrack and charge from BaseTau selection
      • put in as preference rather than flat cut
      • Put in possibility to use mu-mu control region for QCD LepIso measurement.
  • plotQuick -> replace all instances of weight = '1.' to weight = cacheGet( 'tree_weight' )
    • then in the actual hist getter put if weight == None: weight = '1.'

Med Priority

  • Calculate/Implement experimental systematic uncertainties
  • Make 1fb-1 estimates:
    • Mass Fitting (Will)

Low Priority

  • Include Electron Channel
  • Look at Shao Mass
  • Invariant Mass Optimisation:
    • Cut on recoil jet to remove shoulder in signal mass distribution
  • Get simple counting Number Estimates for InvMass Analysis (David)

  • Optimisation:
    • Write full optimiser tool that:
      • Takes in a list of variables
      • Optimises cuts on vars by Minimising ∆N(sig)/N(sig), where:
        • N(sig) = N(data) - Nest(bkg)
        • ∆N(sig)^2 = ∆N(data)^2 + ∆Nest(bkg)^2 = N(data) + (Nest(bkg)*(∆N/N)sys)^2
      • Orders cuts by effectiveness

Completed Tasks (Move here once done)

  • Fitting (Will)
    • Finish ToyMC implementation
    • Add FittingTools module to TauEffPyTools
    • Fit individual backgrounds and set normalisation ranges in fit (eg. +- 10-20%)
    • Implement Toy MC to evaluate systematic uncertainty from MC sample normalisation
    • Ask EB how to implement direct systematic calculation in fit

  • Get simple counting Number Estimates for VisMass Analysis (David)
  • Make 1fb-1 estimates:
    • Event Counting (David)

  • Optimise Analysis Selection (David)
    • Currently 1/7 signal purity in VisMass, should aim for ~ 1/5
    • Variables to Look at:
      • Lepton Isolation
      • Lepton Pt
      • dPhi( lep, tau )
      • MET significance

  • Start Running on 2011 D3PDs
    • Get 2011 ZToTwoTaus branch running (David)
      • Make simple comparison to 2010
      • Fix any major problems with code (see below)
    • Data:
      • Run over periods A-D (period E has LAr FEB problem)
      • Calculate total Lumi (use Lumi calc tool)
    • run over mc10a skims here: /home/data/ZtautauLHAnalysis/mc10a

  • Changes to be implemented:
    • General Config:
      • Set Data Lumi to that calculated above for period A-D. (i think this is in samples.py)
    • Pileup Reweighting:
      • only on MC
      • remove vertex reweighting
      • copy over MC and data lumi block hists (from above), put all period A,B,D,E and A-D in.
      • Allow input hists to be configurable in code, and set to run on period A-D.
    • Muons:
      • Problem in TRT cleaning:
        • TRT has an int->float conversion error: nOut/n < 0.9, but nOut/n = 0 since both ints
        • MuonSelectorTool.cxx lines: 194 & 197, should be (float)nOut/(float)n<0.9
      • Efficiency corrections:
      • Isolation corrections:
        • There are no recommended iso corrections for muons yet. Apparently the iso dist matches well after pileup reweighting. We should check this.

-- WillDavey - 20-Apr-2011

Edit | Attach | Watch | Print version | History: r34 < r33 < r32 < r31 < r30 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r34 - 2011-10-30 - WillDavey
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Main All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback