This twiki aims to be a resource of information and a tutorial of the SPyRoot package used by the AtlasCATSUSY group at CERN


The aim of this page is to provide a basic introduction to SPyRoot data analysis package used by the AtlasCATSUSY group at CERN.

The code of SPyRoot is currently located under the following afs directory: /afs/cern.ch/atlas/project/cern/susy1/SPyRoot/

  • the code is maintained using SVN

The part of the code that is common to all susy analysis is located under this directory: /afs/cern.ch/atlas/project/cern/susy1/SPyRoot/common_susy11b/

Main Selection Steeps

Preliminary steeps:

This section provides a set of instructions to be able to quickly produce basic plots using already made cat-susy ntuples.

  1. Request access to /afs/cern.ch/atlas/project/cern/susy1/ by contacting one of the AtlasCATSUSY organisers
  2. login to lxplus :
    • ssh -Y lxplus.cern.ch
    • cd /afs/cern.ch/atlas/project/cern/susy1/
  3. Under this directory make your own copy of SPyRoot by following this steps :
    • cp -rp test_common_susy11b your-name_common_susy11b
    • for example : cp -rp test_common_susy11b Carlos_common_susy11b
    • cd your-name_common_susy11b
  4. Edit setup.sh file and specify your local directory as workdir, change this line :
    • export WorkDir=/afs/cern.ch/atlas/project/cern/susy1/SPyRoot/your-name_common_susy11b
  5. It is possible to define new configuration files (Cut_Files) under dir "Config"
  6. Should be possible to run spyroot from command line
  7. Select the configuration file , execute_&_load step0 ntuple processing (step0 = trigger_selection ):
    • PyROOT> cutSet='myConfigFile'
    • PyROOT> execfile('LoadStep0.py')

Data selection steeps:

  1. Make pickled files from D3PD (in castor), this is known as step -1: *./setupNtuples.py
    • [delete old pickle files to re-do]

  1. Split samples (at step -1) into sub-samples:
    • ./splitSample.py -f files cutset step sample1 sample2 ...
    • [no sample specified: all !]
    • ./splitSample.py -f 5 myConfigFile -1 JetTauEtmiss+
    • ./splitSample.py -f 10 myConfigFile -1 Wenu+
    • ./splitSample.py -f 10 myConfigFile -1 Zee+
    • [ delete old pickle files to re-do]

Step 0:

  1. Trigger selection
  2. Run : ./submitStep.py -c myConfigFile 0 JetTauEtmiss+
  3. Run : ./submitStep.py -c myConfigFile 0 Zee+
    • [overwrite castor ntuples, and pickle files]
    • [overwrite pickle files in afs group space]

  1. Merging of sub-samples into one can be done at any step for example:
    • ./mergeSamples.py [cutSet] [step] [sample] [numSamples]
    • ./mergeSamples.py myConfigFile 0 JetTauEtmiss00180400

Step 1:

  1. Run : ./submitStep.py -c elICHEP_v3 1 T1+ L1Calo00158801 MuonswBeam00158801
    • Usage: submitStep.py [-q ] [-c ] [sample1] [sample2]

Step 2 :

  1. Run : ./submitStep.py -c elICHEP_v3 2 T1+ L1Calo00158801 MuonswBeam00158801
    • [ as above ]

  1. for MC: merge sub-samples into one again:
    • ./mergeSamplesStep2.py elICHEP_v3 T1
    • ./mergeSamplesStep2.py muICHEP_v3 T1

  1. update lumi information:
    • ./updateLumi.py muICHEP_v3 MuonswBeam L1_MU6
    • ./updateLumi.py elICHEP_v3 L1Calo L1_EM5

Now you are ready to analyse cat-susy ntuples

  1. Get a data sample handler object :
    • PyROOT> a=Data.Samples['L1Calo00158582']
  2. Print efficiencies from all algorithms:
  3. Print efficiency from algorithms that contain the word 'Cut' in their name :
    • PyROOT> a.PrintCutEff(cutReq='Cut')
(usually, the last few lines from this printout show the same efficiency, this is due to virtual cuts )

tips and tricks:

  1. you can get help by doing :

which gives Help on method PrintCutEff in module Sample:

PrintCutEff(self, cutNames=[], UseEventWeight=True, applyAllPreviousCuts=True, cutReq='') method of Sample.Sample instance

or you can do a.[TAB_completion] which gives you the possible methods of object a:

  1. PyROOT> a.[TAB_completion]
a.AddDirectory a.Files a.Lumi a.__dict__ a.className
a.AddFiles a.GetCutEff a.NormalizeHist a.__doc__ a.description
a.AddFilesToChain a.GetCutEffErr a.Print a.__getattribute__ a.eventWeightBranch
a.AddFriendSample a.GetCutEffNumbers a.PrintCutEff a.__hash__ a.eventWeightVar
a.AddStatsAfterAlgo a.GetEntries a.Project a.__init__ a.genEvents
a.AddStatsBeforeAlgo a.GetExpectedEvents a.RegisterStatsAlgo a.__mergeSampleResults__ a.initChain
a.Chain a.GetLumi a.RunNumber a.__module__ a.listPickleVars
a.ChangeNtupleDir a.GetStatistics a.Scan a.__new__ a.moduleName
a.CloneTree a.GetTTreeCutEff a.Statistics a.__reduce__ a.name
a.Copy a.GetTTreeCutEffErr a.UnsubscribeStatsAlgo a.__reduce_ex__ a.pickle
a.Dir a.GetTTreeCutEffNumbers a.Weight a.__repr__ a.pickleFileName
a.Draw a.GetWeight a.XSection a.__setattr__ a.reg_stat_algos
a.Draw2D a.GetWeightedEntries a._Sample__pathToPickleFile a.__str__ a.resultTObjects
a.Draw2D_UB a.GetXSection a.__baseStr__ a.__weakref__ a.results
a.Draw3D a.Leaves a.__class__ a.addSample a.tags
a.Draw3D_UB a.LiveTime a.__delattr__ a.analyses a.unpickle

It does not work If you tab from an object that you get from a list, example this will not work :

  • Data.Samples[L1Calo00154815'].[TAB_completion]

You can get a TTree (TChain) by doing :

  • a.Chain
    • (doesnt work for combined samples):

So you can do (for example):

  • PyROOT> a.Chain.Draw("myEl_p_T")
  • PyROOT> a.Chain.Scan("myEl_p_T:myEl_eta:myEl_phi","abs(myEl_eta)<1.0")

********* ******** ************* ********** *********
Row Instance myEl_p_T myEl_eta myEl_phi
******** ******** ************* ************ ********
3 0 10566.992 -0.329333 -0.235800
10 0 11648.953 -0.548596 0.7006354
11 0 18206.357 -0.608094 1.3962770
13 0 13826.779 0.7104663 0.4839378
... ... ... ... ...

to find out what variables are available you can do:

Variables in ntuple

from the same command you will also get group of variable names (mu_staco_E , mu_staco_allautthor , .... are also variables):

* mu_staco_ * : E allauthor author barrelSectors bestMatch charge d0_exPV endcapSectors energyLossErr energyLossPar eta etcone20 etcone30 etcone40 hastrack isCombinedMuon isLowPtReconstructedMuon isStandAloneMuon m matchchi2 matchndof n nBLHits nBLSharedHits nCSCEtaHits nCSCEtaHoles nCSCPhiHits nCSCPhiHoles nGangedPixels nMDTBEEHits nMDTBIHits nMDTBIS78Hits nMDTBMHits nMDTBOHits nMDTEEHits nMDTEIHits nMDTEMHits nMDTEOHits nMDTHits nMDTHoles nOutliersOnTrack nPixHits nPixHoles nPixSharedHits nRPCEtaHits nRPCEtaHoles nRPCLayer1EtaHits nRPCLayer1PhiHits nRPCLayer2EtaHits nRPCLayer2PhiHits nRPCLayer3EtaHits nRPCLayer3PhiHits nRPCPhiHits nRPCPhiHoles nSCTHits nSCTHoles nSCTSharedHits nTGCEtaHits nTGCEtaHoles nTGCLayer1EtaHits nTGCLayer1PhiHits nTGCLayer2EtaHits nTGCLayer2PhiHits nTGCLayer3EtaHits nTGCLayer3PhiHits nTGCLayer4EtaHits nTGCLayer4PhiHits nTGCPhiHits nTGCPhiHoles nTRTHighTHits nTRTHighTOutliers nTRTHits nTRTOutliers nucone20 nucone30 nucone40 phi phi_exPV pt ptcone20 ptcone30 ptcone40 px py pz qoverp_exPV theta_exPV trackd0 trackfitchi2 trackfitndof trackphi trackqoverp tracktheta trackz0 z0_exPV

This tries to print vector quantities on 1 line (as shown above for the staco muons) by splitting on the first _

You can also draw from SPyRoot itself (ie. without going through the TChain). this is good when you have weighted events, and also for combined samples). the method is:

  • PyROOT> help( a.Draw )
    • Draw(self, expression, cut='1', opts='', bins=100, min=0, max=1000, HistName='Hist', Norm=1000.0, maxLumi=-1, DoDraw=True, title='', UseEventWeight=True, defaultSystematic=0.0, systematics={}, DontAllowNegativeBins=False, UseSampleWeight=True) method of Sample.Sample instance
      • The above is samples's Draw method to draw any TTree variable of this sample.
      • maxLumi: if you want to plot only a sub-sample of the dataset, use this option. Note that if maxLumi > 0, then Norm is set to be -1 (ie. histogram is NOT normalized)
        • maxLumi = -1 : run over full sample, apply normalization Norm
        • maxLumi > 0 : run over sub-sample, do not normalize histogram

For example you can do:

  1. PyROOT>hist=a.Draw("myEl_p_T/1000.","1","",100,0,100.,Norm=-1)

For data samples the sample knows its lumi. which you can get by doing :

  1. PyROOT> a.GetLumi()
    • 16021.464999999998 *this is in microbarn (so divide by 1e6 for picobarn).

Combined Samples :

you can combine together different samples to make a combined sample (eg. All data runs into the full data set, or J0,J1,J2,J3,J4,J5,J6 MC pythia jet samples into a total jet background).

you can get this combined samples by doing :

  1. PyROOT> AddCombinedSampleReg(CombinedSampleName='combData', SampleExp='____', Tags='data' )
    • here you add together all samples with the tag 'data' (which all data samples have).

or alternatively you can do:

  1. PyROOT> listOfSamples=['L1Calo00155280','L1Calo00155669','L1Calo00158801']
  2. PyROOT> AddCombinedSample('combData',Data, listOfSamples)

for data the combined sample have all the events in the sub samples and also the combine lumi of these samples. but this is more useful for MC samples where you can combine samples with different weights, by doing for example:

  1. PyROOT> AddCombinedSample('jets',Data,['J0','J1','J2','J3','J4','J5','J6'])

then you can make plots for the combines sample jets like:

  1. PyROOT> hist = Data.Samples['jets'].Draw("myJet_p_T/1000.","1","",100,0,100,Norm=1)
    • here the plots are normalized to 1pb-1 of data (thats what Norm=1 means!).

another nice thing you can use is the SampleHandlers compare function:

  1. PyROOT> hists = Data.Compare(['jets','J0','J1','J2','J3','J4','J5','J6'],"myJet_p_T[0]/1000.","1",100,0,100.,Norm=1)
    • so this makes the same plot for each of the given samples and Draws them on top of each other.

Some info on our specific ntuples:

The ntuples that are loaded after step2 have had:

  • GRL, MET cleaning cuts, Vtx cuts, and require exactly 1 selected isolated lepton with pt>10GeV to be in the event (after overlap removal etc...).
  • In Step2 we apply some 'virtual' cuts which are applied and feature in the cut statistics - but events get into the ntuple even if they fail these.

these can then be applied by selecting on the bits:

details of the cuts can be seen in these configuration files (used in step0,1,2):

  • /afs/cern.ch/user/j/jboyd/scratch1/susy_grp/SPyRoot/susy10_prod/Cuts/elICHEP_v3.py
  • /afs/cern.ch/user/j/jboyd/scratch1/susy_grp/SPyRoot/susy10_prod/Cuts/muICHEP_v3.py

for cuts specific to El or Mu mode - or in:

  • /afs/cern.ch/user/j/jboyd/scratch1/susy_grp/SPyRoot/susy10_prod/Cuts/ICHEP_v3.py

The ntuples contain vectors for selected objects:

ntuple object

For the El,Mu and Jets these contain some of the basic 4-vector information and the index of the object in the default list (also stored in the ntuple) - this index is called eg. myMu_idx.

  • PyROOT> ds['MuonswBeam00159041'].Chain.Scan("myMu_p_T:myMu_eta:mu_staco_pt[myMu_idx]:mu_staco_eta[myMu_idx]")

******** ********* ************* ************Sorted ascending ******** *********
Row Instance myMu_p_T myMu_eta mu_staco_ mu_staco_
******** ********* ************** *********** *********** ******
2 0 10201.325 -0.826803 10201.325 -0.826803
0 0 10686.618 0.0948067 10686.618 0.0948067
1 0 11085.493 0.1847464 11085.493 0.1847464

The above is printing the pt and eta of selected muons in 2 ways but getting the same values.

Major updates:

-- CarlosChavez - 20-Jul-2010

%RESPONSIBLE% CarlosChavezBarajas
%REVIEW% Never reviewed

Edit | Attach | Watch | Print version | History: r7 | r5 < r4 < r3 < r2 | Backlinks | Raw View | Raw edit | More topic actions...
Topic revision: r3 - 2011-06-06 - CarlosChavezBarajas
    • 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 & 2008-2020 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback