BoostedRazorAnalysis

Run II Search for Supersymmetry using Razor variables in the boosted regime

Checkout recipe

export SCRAM_ARCH=slc6_amd64_gcc530
cmsrel CMSSW_8_0_20
cd CMSSW_8_0_20/src
cmsenv
git cms-addpkg RecoLuminosity/LumiDB
git clone https://github.com/jkarancs/BoostedRazorAnalysis
scram b -j 20

Run the Analyzer

When running for the first time, run the setup script in order
to mount eos on lxplus, create softlinks and create/update filelists

cd BoostedRazorAnalysis/Analyzer
python scripts/setup.py

Before running check/set the options for the Analyzer in settings.h

More info about Analyzer:

• The Analyzer program is basically an event looper, histograms/methods/cuts etc. are defined in Analysis_[Name].h
• All settings (printed also while running) are defined in settings_[Name].h it also includes the Analysis_[Name].h
• common methods (for all analyses/studies) are defined in common/AnalysisBase.h
• The cross-sections and total weight is taken straight from ntuple files
• Reweighting and systematics weight methods are also given in AnalysisBase
• Additionally there's an option (in settings_[Name].h) to save a skimmed TTree with same content for the selected events
• counts are saved for all common and specific analysis cuts in the order they are defined

Run your anaylsis code with

make clean; make Analyzer
./Analyzer <output filename> <filelist.txt or root file(s) ... >
eg:
./Analyzer Bkg_TTJets_madgraph.root filelists/backgrounds/TTJets_madgraph.txt

There is a py script to run the Analyzer to run on filelists of datasets with lot of options to use (use option --help)

python scripts/run_all.py --help

Run a quick (1/Nth events) interactive test, using 4 cpus on all/selected samples

python scripts/run_all.py --full --nproc=4 --quick=100 --run
python scripts/run_all.py --full --nproc=4 --quick=100 --run filelists/data/*.txt

Run a full analysis on all/selected samples in parallel interactive jobs, using 4 processors

python scripts/run_all.py --full --nproc=4 --run
python scripts/run_all.py --full --nproc=4 --run filelists/data/*.txt

If you are on lxplus, it is highly recommended to run full analysis on the batch First a quick test on a faster queue N.B: you need to generate a temp filelist for split jobs, this is a bit slow, so next time you can reuse previous list

python scripts/run_all.py --batch --queue=8nm --quick=100 --sleep=0 --nevt=6000000 filelists/backgrounds/TT_powheg-pythia8_ext3_reHLT.txt
python scripts/run_all.py --batch --queue=8nm --quick=100 --sleep=1 --useprev --run filelists/backgrounds/TT_powheg-pythia8_ext3_reHLT.txt

Then, do a full analysis on the recommended default 1nh queue, default 3s sleep between submits

python scripts/run_all.py --full --batch --sleep=0 --nevt=2000000
python scripts/run_all.py --full --batch --useprev --run
python scripts/run_all.py --full --batch --useprev --run filelists/data/*.txt

Run systematics (first, set varySystematics = true in settings.h)
when running systematics, make sure you define one histogram for each variation syst.index in the Analyzer.cc files can be used to select a histogram from a vector

You can remake the systematics file with:

python scripts/write_syst.py <nline>

Contact Information

Janos Karancsi (janos.karancsi@cern.ch)