Presentation is loading. Please wait.

Presentation is loading. Please wait.

Max BaakNikhef PAT meeting, 24 May 1 HighPtView ntuples in TV modular analysis.

Similar presentations


Presentation on theme: "Max BaakNikhef PAT meeting, 24 May 1 HighPtView ntuples in TV modular analysis."— Presentation transcript:

1 Max BaakNikhef PAT meeting, 24 May 1 HighPtView ntuples in TV modular analysis

2 Max BaakNikhef PAT meeting, 24 May 2 Situation in TVma framework  Useful if TVma framework can process other ntuple formats than just TopView  Specifically, for now: EventView type ntuples  This talk: HighPtView.  Many generic modules useful to have working for any analysis  e.g.: selection modules, QA modules.  Very nice if these work for any ntuple format  Straightforward to add new tree interfaces in TVma  Thanks to Manouk and Wouter.  TVma modules tailored to TopView ntuple format (obviously)  specific use of TV tree names (FullReco0 and TruthAll0)  specific use of TV variable names

3 Max BaakNikhef PAT meeting, 24 May 3 Few words on HPTV  Reasonable starting point for nearly any new physics analyses.  Specifically: non-top analyses.  “Consistent” ntuples:  Preselection, overlap-removal applied  Trigger info available (though currently incorrect)  Combinations of Muid/Staco and TauRec/Tau1p3p  Four jet algorithms: Cone0.4, Cone0.7, Kt0.7 (, Kt0.4 when fixed)  Contain 3 trees: FullRec0, Truth0, FastSim0  “Limited” full truth information  See: https://twiki.cern.ch/twiki/bin/view/Atlas/HighPtView for more info  HPTV ntuples already produced centrally for all requested MC, starting from release 12.0.6.5  dq2_ls *HPTV*

4 Max BaakNikhef PAT meeting, 24 May 4 Extending TVma  Manouk & Wouter: TVma framework can already process TV FastSim ntuples.  Contain a FastSim0 tree instead of a FullReco0 tree.  But: analysis modules referencing FullReco0 tree can conditionally access every FastSim0/FullReco0 variable, depending on “FastSim” flag  Workable solution for just TV ntuples... ... but addition of more interfaces gets rather cumbersome  e.g. HPTV FullRec0 and FastSim0 tree  SUSY, HiggsView?  Not managable solution in the long run (imho).  Don’t want switches at every variable...  Work-around easily leads to unnecessary code-duplication.

5 Max BaakNikhef PAT meeting, 24 May 5 Overlapping variables and adaptation  Many overlapping basic variables between TopView and HighPtView FullReco/FastSim trees.  Some overlapping variables in Truth(All) trees.  TV and HPTV structures quite different.  HPTV dumps selected list of tru particles – but not full truth info.  No need to adapt all top-analysis specific modules to HPTV  Bound to use TopView anyhow (for now)  Focussed on adaptation of selection and QA modules.  Good starting point for new analyses...

6 Max BaakNikhef PAT meeting, 24 May 6 Adaptation of TVma for HPTV  Created 2 commom base classes:  CommonRecVars for TV and HPTV interfaces to FullReco/FastSim  CommonTruVars for TV and HPTV interfaces to Truth(All) TopView trees: FullReco0 FastSim0 Truth0 TruthAll0 HighPtView trees: FullRec0 FastSim0 Truth0 CSCCommomRecVars: RecBase CSCCommomTruVars: TruBase

7 Max BaakNikhef PAT meeting, 24 May 7  Created 2 commom base classes:  CommonRecVars for TV and HPTV interfaces to FullReco/FastSim  CommonTruVars for TV and HPTV interfaces to Truth(All)  Both classes contain (dummy) virtual functions to access variables (we want to be) common between TV/HPTV and FullReco/FastSim.  List contains all basic variables: (tru) kinematic, MET, eventWeights, etc. (All variables used in TVma.)  e.g. El_px, Mu_eta, MET_Final_Et  See CSCCommonRecVars.h or CSCCommonTruVars.h for full lists  Easy to add more common variables to list...  And, when using TopView (HighPtView) ntuples:  CommonRecVars *RecBase = FullReco0; // (HFullRec0);  CommonTruVars *TruBase = TruthAll0; // (HTruth0); Adaptation of TVma for HPTV

8 Max BaakNikhef PAT meeting, 24 May 8 Adaptation of TVma for HPTV  For example, you can now do:  RecBase  vEl_eta(idx)  FullReco0  vEl_eta(idx)  (in case of TV AtlFast MC)  FastSim0  vEl_eta(idx)  (in case of HPTV)  HFullRec0  vEl_eta(idx)  (in case of HPTV AtlFast MC)  HFastSim0  vEl_eta(idx)  RecBase  get_MET_Final_Ex()  FullReco0  MET_Final_Ex  (in case of HPTV)  HFullRec0  MET_Final_Ex  RecBase  vPJet_px(idx)  FullReco0  vPJet_px(idx)  (in case of HPTV)  HFullRec0  vJet_C4_px(idx)  RecBase  vEl_etcone40(idx)  See next slide...  In selection and QA modules, straight-forward replacement of instances “FullReco0  ” by “RecBase  ” to get them to work for both TV and HPTV and/or FastSim/FullReco MC:  EventCounter, FillAODVectors, ElectronSelection, MuonSelection, JetSelection  BTagQA, ElectronQA, EtmissSelQA, JetQA, MuonQA non-overlapping variable

9 Max BaakNikhef PAT meeting, 24 May 9 “Troublesome” variables to convert  HPTV ntuples contain “limited” full truth information  Most truth particles stored in “Sp_” container, still...  No parent- & daughter barcodes included  no browsing through decay trees possible  I’ve brought this up with Amir F – ROOT but, fixed in near future  Variables I’ve come across in TV fullreco code, not in HPTV...  El_author, El_NBLayerHits, El_NPixelHits, El_NSCTHits, El_NTRTHits, El_NHighThresTRTHits, Mu_author  Variables not in FastSim trees...  El/Mu_etcone*, Mu_nucone*  In TV fastsim or HPTV ntuples these currently return dummy values and a warning. E.g. RecBase  vEl_vNTRTHits(idx)  “CSCCommonRecVars: WARNING: El_NTRTHits not defined in ntuple. Using dummy value: -999”

10 Max BaakNikhef PAT meeting, 24 May 10Conclusion  What has changed for your (top) analysis modules? Nothing!  Code entirely backwards compatible  TVModularAnalysis V00-00-02 can read both TV and HPTV ntuples  Set “NTupleType” flag to 2 to select HPTV [set to 1 (TV) by default].  Set “FullTru” flag to 0  When changing selection or QA modules, please use calls to RecBase (or TruBase) to ensure code works for both TV/HPTV and FullReco/FastSim ntuples.  If no common cut/variable available:  Add dummy variable to RecBase (or TruBase), or  use “NTupleType” and/or ”FastSim” flag  Easy to add (basic functional use of) other EventView style ntuples  SusyView, HiggsView, etc.  Will add section in the README file describing howto.

11 Max BaakNikhef PAT meeting, 24 May 11Homework  In your specific reconstruction analysis module(s), try changing every instance of “FullReco0” to “RecBase”.  Change calls to integers and double (not a vector):  e.g. FullReco0  PJet_N to RecBase  get_PJet_N()  When running on fully reconstructed TV ntuples  functionality should be unchanged.  Now try running on... 1TV AtlFast MC, using flag “FullTru=0”, and see what happens... 2HPTV fully reco’d MC, using “NTupleType=2”, and see what happens... 3HPTV AtlFast MC, using “NTupleType=2” and “FullTru=0”, and see again...

12 Max BaakNikhef PAT meeting, 24 May 12Prospects  Current TVma framework quickly adaptable for other EventView style ntuples.  Looks like pAOD may become default derived physics data format  Still in development, significant changes for TVma likely.  TVma framework converts El, Mu, Jet in TLorentzVector objects with associated index.  Index allows back navigation to full AOD info in ntuple  Idea from Wouter: “Universal Portable Analysis”  Add additional variables to these extended TLorentzVector objects. Make analysis modules entirely independent of (specific) ntuple interface(s).  Easy for writing out / reading in subskims  Requires migration of all analysis modules


Download ppt "Max BaakNikhef PAT meeting, 24 May 1 HighPtView ntuples in TV modular analysis."

Similar presentations


Ads by Google