Skip to content

noamhod/Tau3mu.8TeV.NTUPanalysis

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

20 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Tau3mu.8TeV.NTUPanalysis

Tau3mu 8TeV NTUPanalysis based on NTUPmaker output

Analysis flow:

  • to get the trunk: svn co svn+ssh://svn.cern.ch/reps/atlasphys-exa/Physics/Exotic/Analysis/TauToMuX/ThreeMu/NTUPanalysis/trunk NTUPanalysis

  • to tag a version in svn: See the latest tag in: https://svnweb.cern.ch/trac/atlasphys-exa/browser/Physics/Exotic/Analysis/TauToMuX svn cp svn+ssh://svn.cern.ch/reps/atlasphys-exa/Physics/Exotic/Analysis/TauToMuX/ThreeMu/NTUPanalysis/{trunk,tags/NTUPanalysis-XX-YY-ZZ} -m "tagname"

  • need to have flatout.periodall+MC.muons.cuts.analysis.n0.j0.root in this directory

    • can be downloaded from /eos/atlas/user/h/hod/data/flat/ like that: eos cp /eos/atlas/user/h/hod/data/flat/flatout.periodall+MC.muons.cuts.analysis.n0.j0.root $HOME/
  • root -b -l -q compile.C --chnl=all[or a single channel name]
    --master=muons[or muid]
    --method=cuts[or mva]
    --jetmode=calib[uncalib, calib, JESUP, JESDWN, JERUP, JERDWN]
    --metmode=uncalib[uncalib, calib, JESUP, JESDWN, JERUP, JERDWN]
    --blinded=yes

    • take a look at:
      • the cutflow
      • figures.cuts.pdf
    • root -b -l -q fit3body.C++("muons","cuts") # or "muid"
      • take a look at the mass fit plots
    • root -b -l -q cutflow.C++("muons","cuts") # or "muid"
      • take a look at the produced tex file and copy the big cutflow table
    • root -b -l -q objectquality.C++("muons","cuts") # or "muid"
      • take a look at the objqa.pdf file
  • root -l ./tmvaClass.C("BDTG")

    • produce the MVA plots...
    • root -l getUL.C
    • look for the minimum...
    • root -l UL.root
    • look at the 3 histos above BDTG score=0.9 and chose the best cut
    • update the cut value in fltvtx.C
  • root -b -l -q compile.C --chnl=all --master=muons --method=mva

    • take a look at:
      • the cutflow
      • figures.mva.pdf
    • root -b -l -q fit3body.C++("muons","mva")
      • take a look at the mass fit plots
    • root -b -l -q cutflow.C++("muons","mva") # or "muid"
      • take a look at the produced tex file and copy the big cutflow table (just the line of the BDT cut)
  • root -b -l -q compile.C --chnl=all --master=muons --method=mvaout (produce optimization input file without cutting on the score)

    • copy mvaout.root to the HistFitter directory on lxplus and run the mva optimization there:
    • scp mvaout.muons.root hod@lxplus.cern.ch:HistFitter/HistFitter-00-00-37/
    • root -b -l -q readmvaout.C++
      • look on the file: readmvaout.pdf (optimization plot is in the last page)
      • write down the best MVA cut
    • choose the 3body mass fit plot corresponding to the best MVA cut
      • change the expected background events and its error in python/MyUserAnalysis.py (verify all other parameters)
      • choose once the case for a downwards fluctuation in Nobs wrt the expectation
      • only after copying the CLs plot, run the same but for an upwards fluctuation
      • run it with: ./run.sh -l
      • write down the values in the fit result
      • look at the CLs result in file: results/upperlimit_cls_poi_Sig_Asym_CLs_grid_ts3.root.eps
  • Proceed to tex build

    • first run: ./getEps.sh cuts[mva] muons[muid]
    • make

Short setup for optimization: * freeze loose cuts * root -b -l -q compile.C --chnl=all --master=muons --method=cuts --jetmode=calib --metmode=calib --mettype=muons --blinded=yes * Can train on a range slightly larger than the actual sidebands but without the blinded area: -- train: [1300,1690]+[1870,2320] -- apply: [1450,1690]+[1870,2290] -- note the b3body bin size (chosen as the resolution) * chose percentage of events (80%) to use form the signal / background passing events and train the BDT with: -- root -l ./tmvaClass.C("1300.","1690.","1870.","2320.","muons","BDTG") * look at the training vs test samples * root -b -l -q compile.C --chnl=all --master=muons --method=mvaout --jetmode=calib --metmode=calib --mettype=muons --blinded=yes * change the min BDT cut and the BDT binnings if necessary (in "const.h") * root -l -b -q bdtroofit.C++ * look on the fit(s) results * scp mvaout.muons.root BDTfitResults.root hod@lxplus.cern.ch:HistFitter/HistFitter-00-00-37/ * ssh hod@lxplus.cern.ch * cd HistFitter/HistFitter-00-00-37; source setup.sh; * root -b -l -q readmvaout.C++(1500.,1700.,1860.,2100.,20.,0.8000,0.9990,0.001,20.28) * scp hod@lxplus.cern.ch:HistFitter/HistFitter-00-00-37/readmvaout.pdf readmvaout.fit.new.pdf * look on the optimization and update the optimal BDT cut in const.h * root -b -l -q compile.C --chnl=all --master=muons --method=mva --jetmode=calib --metmode=calib --mettype=muons --blinded=yes * finalisations: -- root -b -l -q cutflow.C++("muons","cuts") # or "muid" -- root -b -l -q objectquality.C++("muons","cuts") # or "muid" -- root -b -l -q phitest.C++ -- root -b -l -q sidebandsBias.C++ -- root -b -l -q sidebandsDef.C++ -- source runsystsig.sh -- source runRangeSyst.sh dry/full -- root -b -l -q totalSyst.C -- root -b -l -q allFitSystAuto.C++ -- source copyRangeSyst.sh * login lxplus -- cd HistFitter/HistFitter-00-00-37/ -- source setup.sh -- root -b -l -q readmvaout.C++(1450.,1690.,1870.,2290.,50.,0.9000,0.9950,0.005) * back on mac: -- scp hod@lxplus.cern.ch:HistFitter/HistFitter-00-00-37/Optimization2D.1450-1690-1870-2290..pdf -- scp hod@lxplus.cern.ch:HistFitter/HistFitter-00-00-37/Optimization2D.1450-1690-1870-2290..root -- root -b -l -q optimization2Dreplot.C++

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published