Tools to convert the output of MG5 generation with EFT components in readable histos for mkDatacards.py in the Latino framework. This tool allows to create histograms compatible with the new combine models for EFT fitting in the AnalyticAnomalousCoupling repo.
Useful links:
-
mkDatacard.py: https://github.com/latinos/LatinoAnalysis/blob/master/ShapeAnalysis/scripts/mkDatacards.py
-
example folder: https://github.com/amassiro/PlotsConfigurationsCMSDAS2020CERN/tree/master/ControlRegions/DY
-
D6EFT study for generation: https://github.com/UniMiBAnalyses/D6EFTStudies
-
Latino tutorial CmsDas h->ww : https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolCERN2020LongHWW
-
AnalyticAnomalousCoupling combine model : https://github.com/GiacomoBoldrini/AnalyticAnomalousCoupling
cmsrel CMSSW_10_6_4
cd CMSSW_10_6_4/src/
cmsenv
If you want to generate samples clone the D6EFT repo and follow the instruction provided in the link above.
git clone [email protected]:UniMiBAnalyses/D6EFTStudies.git
Setup the latino framework (remember to clone/source with the ssh key connected to git)
git clone --branch 13TeV [email protected]:latinos/setup.git LatinosSetup
source LatinosSetup/SetupShapeOnly.sh
scramv1 b -j 20
cd LatinoAnalysis/Tools/python
cp userConfig_TEMPLATE.py userConfig.py
# edit the userConfig.py so the paths correspond to a directory where you have write access (this will be used to write log files)
cd ../../../
scram b
Finally clone this repo. The implementation of this framework is independent both from the generation step and from the latinos framework so you can place it wherever you want.
git clone [email protected]:GiacomoBoldrini/D6tomkDatacard.git
This framework works if you can provide it with samples. So you need to generate them first. There are a bunch of them 1M event under /afs/cern.ch/user/g/govoni/myeos/samples/2019_EFT(restricted access)
The routine is specified in a .cfg under the cfg folder. It has many sections.
The generalsection specifies standard naming conventions for the process under study and the output folder/file names in which to store results.
The ntuples section specifies the folder with the ntuples. It can be a list of folders, files will be picked smartly using the glob module.
In the eft section one can specify the operators of interest. If they are more than one the full interference pattern will be saved (planning in the future to save different scenarios for example the single-operator/s one).
For example if two operators are specified you will get a two operator datacard. The section model is crucial as it defines how histos will be combined. It follows the nomenclature of AnalyticAnomalousCoupling.
The model can be chosen from the following EFT,EFTNeg,EFTNeg-alt. It can be a list and all the output shapes will be saved in their own folders.
Section cuts,supercuts are work in progress and do nothing. The variables section specifies the tree names of the variables you want to create shapes for. You are suppposed to give a list of bins (one for each variable) and a list of lists ([xmin:xmax],[],..) with the x-axis ranges in order to build histos.
Sections starting with d_ creates some dummy files needed for mkDatacards.py. You can switch on and off their creation by changin the boolean field makeDummy under each section.
They work but they are not fully reliable due to the enormous flexibility of the Latinos framework. Check and modify them if you have to add backgrounds or different stuffs.
Folders under the ntuples/folder config section do not need to be sorted. However the files should have a definite convention in order for the algorithms to pickup the ones from the right process, component and operator. File ntuple names should have the following syntax:
folder + "/*_" + process + "_" + op + "*.root"
folder + "/*_" + process + "_{}_{}_".format(op[0], op[1]) + "*.root"
process may be something like "SSWW" and should match the one you are giving under the config section general/sample. op is the operator name, should match the ones gave in the config section eft/operators. The order of op[0],op[1] is not important, the script will automatically detect the combination.
Lastly the file name should specify if it contains events simulating the SM, linear, quadratic or interference components. These will have the postfix
SM,LI,QU,IN respectively (no operator specification for the SM sample).
Some examples of file names:
ntuple_OSWW_SM.root #Opposite sign WW process, SM component
ntuple_OSWWQCD_cHl1_QU.root # Opposite sign WW QCD quadratic component of op cHl1
ntuple_SSWW_cHq3_cll1_IN.root # Same sign WW interference term between cHq3 and cll1
ntuple_SSWW_cHWB_LI.root # Same sign WW linear term of operator cHWB_LI
One root file must have a main tree with ntuples named as the file name minus ntuple_ such as OSWW_SM for the first example above. The root file should also have one auxiliary root histogram (TH1F) containing important information in order to weight histograms. The information needs to be encapsulated in the bin contents as follows (no underflow or overflow involved here):
- 1st bin: cross_section
- 2nd bin: total sum of the weights
The histos normalization is computed as: (question: shouldn't we multiply by the weights of the events selected by cuts?)
cross_section * 1000. * luminosity / (sum_weights_total)
The luminosity is taken from the config file and has to be in pb
This is a collection of functions that creates dummy configuration files for mkDatacards.py. Not reliable but a good starting point for more complex use-cases.
This is the main script. It reads the ntuples, make histograms according to the specified model and save them in a way mkDatacards.py can read them. It only expects a config file input.
cd D6tomkDatacard
python mkDCInputs.py --cfg cfg/SSWW_to_mkDat.cfg
cp -r SSWW_to_Latinos/ where_mkDatacard_can_see_it/
cd where_mkDatacard_can_see_it/SSWW_to_Latinos/EFTNeg
mkDatacards.py --pycfg=configuration_SSWW_EFTNeg.py --inputFile=rootFile/cW_cHWB.root
Inspect the datacard folder. It has a subfolder with you process name and many subfolders with the variables you specified from the ntuples.
Each subfolder has a datacard with the EFT components according to the model you chose and to the AnalyticAnomalousCoupling nomenclature (e.g. lin_, quad_ sm_lin_quad_ etc...).
You will find the shapes in the shape folder.
You can also use mkPlot.py, pay attention to the grouping in the cfgfile.
mkPlot.py --pycfg=configuration_SSWW_EFTNeg.py --inputFile=rootFile/cW_cHWB.root
You should see the histograms stacked for the model you specified. If you want to merge the contributions you can act on the grouping options in the cfg file,
before mkDCInputs.py
- Add cuts smartly
- Optimize some sections
- Skim the cfg file
- Add "running on batch" option
- Add intelligent ways of generating AC-DC combine histos for different samples simultaneously