The analysis part is based on ntuples made from

https://github.com/albertobelloni/UMDNTuple

To run the analyzer, first do

git clone https://github.com/albertobelloni/WG_Analysis.git
cd WG_Analysis/

If you are on bash:

source setup.sh

or if you are on tcsh:

source setup.csh

If this is your first time to run the analysis, compile the core first:

cd Analysis/TreeFilter/Core
make

Then modify the script in RecoResonance:

cd ../RecoResonance/scripts/

change the base, output_base, and options.PUPath in scheduler.py if necessary. (By default the base should be already pointed to the correct ntuples. And the PU files should be correct as well. So in principle you only need to modify the output_base to your directory.)

Also change the preselections if needed. Here the make_final_mug is called in Line 243-251. Comment out these lines and uncomment other preselections you want.

After setting up the config file

cd ..

If you want to submit jobs, do

python scripts/scheduler.py

If you want your jobs to run locally, do

python scripts/scheduler.py --local

After finishing, check if all events in the input files have been processed by

python scripts/scheduler.py --check

More information on the Twiki. Try search 'how to process ntuples'

For the MET filter, it's under

WG_Analysis/Analysis/TreeFilter/RecoResonance/scripts/Conf.py

Line 211-213 or Line 263-265

The three lines will pass the cut ids to FilterMET and select events passing these 7 flags, recommended by the MET group (Twiki).

The filter id and name map can be found in the ntuples from UMDNTuple/FilterInfoTree

If you need to apply the MET filter, simply copy these three lines to the pre-selection function, like in make_final_elg or make_final_mug.

After processing, your new ntuples should contain about 20 branches, such as Flag_BadPFMuonFilter, and if everything is correct, the 7 flags should always be 1 in the new ntuples, i.e.,

Flag_HBHENoiseFilter

Flag_HBHENoiseIsoFilter

Flag_globalSuperTightHalo2016Filter

Flag_EcalDeadCellTriggerPrimitiveFilter

Flag_goodVertices

Flag_BadChargedCandidateFilter

Flag_BadPFMuonFilter

After finishing the above, please be careful if you are using inclusive and binned MC samples, you might need to run FilterOverlap to remove overlapping between different MC samples,

cd WG_Analysis/Analysis/TreeFilter/FilterOverlap/scripts/

and modify the base in scheduler.py to the your output.

If this is your first time to run FilterOverlap, please

cd WG_Analysis/Analysis/TreeFilter/FilterOverlap/
mkdir obj

Otherwise it would go into problem when compiling. Finally,

cd ..
python scripts/scheduler.py

In theory the cuts should be the same so you don't need to change other settings.

Go to the Plotting directory first. Open the Makefile, change the OUTPUTMAIN to wherever you want to save the results.

Do

root -l My_double_CB/RooDoubleCB.cc+

to compile Double Crystal Ball. (This function is taken from the HiggsCombine code (Here.))

make signal

to fit the signal templates with double sided crystal ball function and save the fit results (pdfs, vars and norms) into the workspace. After this, a new directory will be created with some root files containing the signal workspaces and fitting plots under the OUTPUTMAIN directory you just set.

After preparing the signal workspaces, you can do

make mcbkg

to prepare the bkg templates for different backgrounds. The default is WGamma only. One could change this in the Makefile. Similarly, new root files with workspaces for backgrounds and fitting plots will be created in the OUTPUTMAIN directory as well.

When the signal and bkg are all done, if you want to run the limit settings, first you have to set up and compile the Higgs Combine code. Create a new directory wherever you like and follow the instructions from

https://cms-analysis.github.io/HiggsAnalysis-CombinedLimit/

(For now we are using the CMSSW_8_1_X version. So you could follow the instructions starting from

https://cms-analysis.github.io/HiggsAnalysis-CombinedLimit/#slc6cc7-release-cmssw_8_1_x

)

(In the last step, instead of scramv1 b, you may want to do scramv1 b -j 6 or scramv1 b -j 8 to increase the number of CPUs used for compiling, in order to speed things up.)

After compiling the HiggsCombine code, change the DIR_Combine in the Makefile to the src location of the CMSSW_8_1_X you just set. Then do

make fits

This will collect the signal and background workspaces, generate the datacard and the code to run combine, and submit the jobs to condor. All these will be saved in the subdirectory under the DIR_Combine/TEST. It might take a few mintues to finish, and after that you will get a few json files with limits for different signal widths and masses, saved in the same directory.

Finally you can load the json file and plot the limits. CombineHarvester has provided some functions and scripts for that. Search 'Plotting' from the above link and you will find some useful information.