DelayedPhoton

Plots and limits for delayed photon analysis

setup

export SCRAM_ARCH=slc6_amd64_gcc530
cmsrel CMSSW_8_1_0
cd CMSSW_8_1_0/src
mkdir -p HiggsAnalysis
cd HiggsAnalysis
cmsenv
git clone https://github.com/cms-analysis/HiggsAnalysis-CombinedLimit.git CombinedLimit
cd CombinedLimit
git fetch origin
git checkout 81x-root606
scramv1 b clean; scramv1 b
cd ../
git clone git@github.com:cms-lpc-llp/DelayedPhotonLimit.git
cd DelayedPhotonLimit
git checkout Limit2018DNN-D
make

If it failed, possibly you need to run singularity first:

export SINGULARITY_CACHEDIR=/tmp/qnguyen/singularity;singularity shell -B /cvmfs -B /storage /cvmfs/singularity.opensciencegrid.org/cmssw/cms:rhel6

to at least after the scramv1 step to compile the Higgs CombinedLimit tool. After that you can run make in normal CMSSW environment without singularity.

Run the analysis step by step

---

1. skim the ntuples (not necessary, just to speed up next steps)

---

cd python
python skim_noBDT.py

skim_noBDT.py uses variables from config_noBDT.py. In particular you need to paste the input list file to fileNameSigSkim variable. The list was created by this command:

    find ~/DelayedPhoton/CMSSW_10_6_12/src/DelayedPhotonID/deployment/output_bothpho_2017_new_signal -name '*.root'

---

2. run the fit and obtain datacards

---

./FitABCD \
/storage/af/user/qnguyen/DelayedPhoton/CMSSW_8_1_0/src/HiggsAnalysis/DelayedPhotonLimit/2018D/skim_DNN/EGamma_Run2018D.root \
/storage/af/user/qnguyen/DelayedPhoton/CMSSW_8_1_0/src/HiggsAnalysis/DelayedPhotonLimit/2018D/skim_DNN/GMSB_L100TeV_Ctau10cm_13TeV-pythia8.root \
"L100TeV_Ctau1cm" \
"signal (L100-Ctau10)" \
3J \
datacard \
no \

or simply run source test.sh.

This will generate the datacard for that signal points for signals of ctau = 10cm, we use v20 (Closure systematics 90%), for other signal points, we use v18 (Closure systematics 2%)

Regarding app/FitABCD.cc:

• The only fitMode that matters is datacard. Other mode like binAndDatacard is used for optimizing the bin, which is not really necessary because it was done before. Feel free to rerun if you like.
• The only categories that matters are 3J (for 2-pho category) and 3J1P (for 1-pho category). The 2J category was legacy from previous analyses, which requires at least 2 jets instead of 3.
• useBDT is always False. We don't use it (legacy from previous analyses). As the result, the output directories are datacards_3J_noBDT and datacards_3J1P_noBDT that you should look at.
• Set _useToy to true for blind analysis. When the data is unblinded, set it to false.

After modifying FitABCD.cc, remember to make again before running the fit with source test.sh.

Once you're sure the code is good, submit it to condor:

cd scripts_condor
python create_list.py # to create the input list of signal, which will produce output mylist2017all.list
python submit_datacard_noBDT_ABCD.py mylist2018all.list 3J datacard

or

python submit_datacard_noBDT_ABCD.py mylist2018all.list 3J1P datacard # for 1-photon category

After condor is done, run python after_condor.sh to unpack the output and save it to combine directory (indicated by appDir variable.

---

3. draw the limit plots

---

Once you run all the signal points and you get the limit trees for all of them, you can plot the 2D exclusion region plot (the one in the paper):

    cd combine
    python cmd_combine_cards_2018D.sh # to combine 1-pho and 2-pho category
    python cmd_combine_cards_2018ABC_D.sh # to combine with 2018ABC, if you have it
    source cmd_combine_cards_Run2.sh # to combine with 2016 and 2017
    python draw_limits_run2.py