draw_pico

draw_pico is a user-level analysis framework for making 1-d histograms, 2-d histograms, efficiency plots, tables, events scans, and statistical models (in the form of combine datacards) from high energy physics n-tuples. Currently supported n-tuple formats include picos and NanoAOD, though new formats can be easily added with a specification text file. It has been used for the the CMS 1 lepton+jets+MET (MJ) SUSY analysis, the HH(4b)+MET analysis, and the CMS H to Z gamma analysis.

Repository inherited from: https://github.com/ald77/ra4_draw

For examples of functionality see: src/core/test.cxx

Setup

Prerequisites: (Note, on UCSB servers, source set_env.sh can be used to fulfill these prerequisites, which is described below in the UCSB server section.)

• ROOT
• (Optional) Scons for building the library. Can also use the compile.py script instead.

Please run the below command to clone the repository

git clone [email protected]:richstu/draw_pico.git
cd draw_pico

Please run the below command to build the libraries.

./compile.py

or

scons

Setup on UCSB servers

If you use the USCB servers, you can use the below commands for the setup. This also sets up the environment for the batch system at UCSB

git clone --recurse-submodules [email protected]:richstu/draw_pico.git
cd draw_pico
source set_env.sh
./compile.py or scons

An alternative method is shown below:

git clone [email protected]:richstu/draw_pico.git
cd draw_pico
git submodule update --init --remote --recursive
source set_env.sh
./compile.py or scons

Plot options explanation

StackType:

• signal_overlay: Background samples are stacked with each other, signals are stacked with each other and overlayed in front. Samples are scaled according to luminosity and cross section.
• signal_on_top: Background samples are stacked with each other, signals are then stacked on top. Samples are scaled according to luminosity and cross section.
• data_norm: Signal and background samples are stacked with each other?? Samples are scaled according to integral of data. Data is plotted as dots.
• lumi_shapes: No samples are stacked. Samples are scaled according to luminosity and cross section.
• shapes: No samples are stacked. Samples are all scaled to 1.
• prop_shape_stack: Background samples are stacked with each other, signal samples are stacked with each other and overlayed. The sum of backgrounds and signals are scaled to the same value. Used to compare the shape of combined signal and background distributions.

YAxisType: Sets the Y axis as selected

• linear
• log

TitleType: Sets the title as selected

• info
• preliminary
• simulation
• simulation_preliminary
• simulation_supplementary
• supplementary
• data

OverflowType:

• none: Overflow bins exist on both sides, but not plotted
• underflow, overflow: The content in the relevant overflow bin is merged into the adjacent plotted bin, making the overflow visible.
• both: Both bins made visible in the fashion described above

BottomType:

• off: No bottom plot
• ratio: A/B plotted on bottom. B is backgrounds if available, signal if no background samples. A is data if available, signal if no data samples.
• diff: A-B plotted on bottom. Same definitions.
• sorb: signal efficiency/Sqrt(background efficiency) after applying lower bound cut at any point along the y axis. This is plotted on the bottom. Works best with signal_overlay stack type, does not work with signal_on_top
• sorb_cut_upper: Same as sorb, but with upper bound cut.

Cutflow table generating tools

Cutflows following the H to Z gamma baseline selection can be automatically generated using the cutflow back-end and input jsons, stored in txt/.

./run/zgamma/cutflow_backend.cxx [cutflow input]

The main input file for creating baseline selection cutflows is cutflow_inputs_datamccombineAN.json. Running with these inputs will create all cutflows necessary for the AN. The other input files can be used for debugging or specific alternative sets of cutflows.

Notes on inputs in json files:

• production_name and short_name are used to create the subdirectories table/production_name/short_name where the cutflows are stored. The subdirectories are also timestamped, so re-running cutflow generation will not replace previous cutflows.
• sample_type should be a string selected from: "sig", "bkg", or "data".
• weight is a string selected from: "1" (unweighted), "lumi" (lumiweighted), "custom" (custom weight based on definition in backend), or "weight" (full weight branch)
• mlly_window is a string that should be selected from: "full", "sideband", and "higgs"
• lepton_type is a list of strings which can be selected from: "el", "mu", and "lep" The remainder of inputs are largely self-explanatory.

There is currently no machinery to combine the separately generated cutflows. This task should be done externally to draw_pico.

Higgsino useful commands

To make datacards and get limits:

To plot overlap between the boosted and resolved analysis:

./run/higgsino/plot_regions.exe #boosted overlap in resolved regions
./run/higgsino/plot_regions.exe --boo #resolved overlap in boosted regions

Write all the datacards:

./run/higgsino/write_kappa_datacards_priority.exe -o datacards/tchihh_onedim/ -m CN -1 -r 1 --unblind --unblind_signalregion
./run/higgsino/write_kappa_datacards_priority.exe -o datacards/tchihh_twodim/ -m N1N2 -2 -r 1 --unblind --unblind_signalregion
./run/higgsino/write_kappa_datacards_priority.exe -o datacards/t5hh_onedim/ -m T5HH -f -1 -r 1 --unblind --unblind_signalregion
./run/higgsino/write_kappa_datacards_priority.exe -o datacards/t5hh_twodim/ -m T5HH -2 -r 1 --unblind --unblind_signalregion

For comparison of signal yields to other tables, remember to use --recomet such that the yield is not averaged with the one obtained using GenMET. To make a datacard for the boosted case, use -t boosted. Use option --unblind to include data. To run on a particular point add, e.g. -p "700_1". For computers with smaller amounts of memory, there may not be sufficient memory to generate all the 2D T5HH datacards at once. In this case, you can split the job in half with the -s flag, which sets a lower/upper bound on the gluino mass.

./run/higgsino/write_kappa_datacards_priority.exe -o datacards/t5hh_twodim/ -m T5HH -2 -r 1 --unblind --unblind_signalregion -s "-2100"
./run/higgsino/write_kappa_datacards_priority.exe -o datacards/t5hh_twodim/ -m T5HH -2 -r 1 --unblind --unblind_signalregion -s 2100

Then to get a limit interactively:

./run/higgsino/scan_point.exe -f datacards/tchihh_onedim/datacard-TChiHH_mChi-700_mLSP-0_Tune_2016,2017,2018_priority1_resolved.txt

To get limits for the full scan in the batch, use the following script to generate batch commands.

./scripts/write_combine_cmds.py --card_dir datacards/tchihh_onedim -m CN
./scripts/write_combine_cmds.py --card_dir datacards/tchihh_twodim -m N1N2
./scripts/write_combine_cmds.py --card_dir datacards/t5hh_onedim -m T5HH
./scripts/write_combine_cmds.py --card_dir datacards/t5hh_twodim -m T5HH

These can be run locally with ./scripts/run_commands.py or submitted to the UCSB batch system with auto_submit_jobs.py. To combine outputs and make the limit plot

cat datacards/tchihh_onedim/scan_point*/limit*txt | sort >> tchihh_onedim_resolved_limits.txt
cat datacards/tchihh_twodim/scan_point*/limit*txt | sort >> tchihh_twodim_resolved_limits.txt
cat datacards/t5hh_onedim/scan_point*/limit*txt | sort >> t5hh_onedim_resolved_limits.txt
cat datacards/t5hh_twodim/scan_point*/limit*txt | sort >> t5hh_twodim_resolved_limits.txt
./run/higgsino/plot_limit.exe -f tchihh_onedim_resolved_limits.txt --drawData -t tchihh_onedim_resolved
./run/higgsino/limit_scan.exe -f tchihh_twodim_resolved_limits.txt -m N1N2 -t tchihh_twodim_resolved --unblind
./run/higgsino/plot_limit.exe -f t5hh_onedim_resolved_limits.txt --drawData -m T5HH -t t5hh_onedim_resolved
./run/higgsino/limit_scan.exe -f t5hh_twodim_resolved_limits.txt -m T5HH -t t5hh_twodim_resolved --unblind

To generate AN tables/plots:

./run/higgsino/an_plot_triggers.exe --unblind --year 2016 --string_options systematic,efficiency,cr
./run/higgsino/an_plot_triggers.exe --unblind --year 2017 --string_options systematic,efficiency,cr
./run/higgsino/an_plot_triggers.exe --unblind --year 2018 --string_options systematic,efficiency,cr
./run/higgsino/an_plot_triggers.exe --unblind --year run2
./run/higgsino/an_plot_selection.exe --unblind --year run2
./run/higgsino/an_plot_backgroundEstimation.exe --unblind --year run2
./run/higgsino/plot_kappas.exe -s search --scen mc -y run2 -o paper_style
./run/higgsino/an_plot_syst_ttbar.exe --unblind --year run2
./run/higgsino/an_plot_syst_zll.exe --unblind --year run2
./run/higgsino/an_plot_syst_qcd.exe --unblind --year run2
./scripts/datacard_syst_utils.py
./run/higgsino/plot_n_minus_1.exe --unblind --year run2 --sample search
./run/higgsino/plot_search_unblind.exe -u -a -o plot_baseline,paper_style,plot_in_btags,plot_in_btags_with_met_split
./run/higgsino/plot_kappas.exe -s search --scen data -y run2 -o paper_style,use_datacard_results,do_zbi --unblind
./scripts/plot_ra4style_results.py

To generate paper plots:

./run/higgsino/plot_search_unblind.exe -u -a -o plot_baseline,paper_style,plot_in_btags,plot_in_btags_with_met_split
./run/higgsino/plot_kappas.exe -s search --scen mc -y run2 -o paper_style
./run/higgsino/plot_kappas.exe -s search --scen data -y run2 -o paper_style,use_datacard_results,do_zbi --unblind
./scripts/plot_ra4style_results.py

To generate supplementary plots:

./run/higgsino/an_plot_triggers.exe -u -o plot,paper_style
./run/higgsino/an_plot_syst_zll.exe --year run2 --unblind -o paper_style
./run/higgsino/an_plot_syst_qcd.exe --year run2 --unblind -o paper_style
./run/higgsino/an_plot_syst_ttbar.exe --year run2 --unblind -o plot_data_vs_mc,paper_style
./run/higgsino/plot_kappas.exe --sample ttbar --scen data --unblind --year run2 -o paper_style
./run/higgsino/plot_kappas.exe --sample zll --scen data --unblind --year run2 -o paper_style
./run/higgsino/plot_kappas.exe --sample qcd --scen data --unblind --year run2 -o paper_style
./run/higgsino/plot_search_unblind.exe --year run2 -u --unblind_signal -o plot_in_btags,plot_in_btags_with_met_split,paper_style,supplementary
./scripts/plot_signal_efficiency.py
./run/higgsino/plot_phase_space.exe
./run/higgsino/plot_supplementary.exe -y run2 -o cutflow
./run/higgsino/plot_supplementary.exe -y run2 -o pie
./scripts/plot_ra4style_results.py --signal

Getting Higgsino cross-section (CN to N1N2) scale factors.

Step 1. Download cross-section files

scp -r lxplus:/afs/cern.ch/user/a/amete/public/EWKGauginoCrossSections_13TeV cross_section

Step 2. Fix bug in script

Set masses, xsecs, xsecUncs to 0 when initializing.

cd cross_section
sed -i 's/std::vector<double>\* masses;/std::vector<double>\* masses=0;/' get_gaugino.C
sed -i 's/std::vector<double>\* xsecs;/std::vector<double>\* xsecs=0;/' get_gaugino.C
sed -i 's/std::vector<double>\* xsecUncs;/std::vector<double>\* xsecUncs=0;/' get_gaugino.C

Step 3. Run script for all mass points.

model "CN" (mixing) or "N1N2" (no mixing).

cd cross_section
../scripts/print_higgsino_cross_sections.py -i /net/cms25/cms25r0/pico/NanoAODv7/nano/2016/SMS-TChiHH_2D -m CN
../scripts/print_higgsino_cross_sections.py -i /net/cms25/cms25r0/pico/NanoAODv7/nano/2016/SMS-TChiHH_2D -m N1N2
../scripts/print_higgsino_cross_sections.py -i /net/cms25/cms25r0/pico/NanoAODv7/nano/2016/SMS-TChiHH_2D -c

Zgamma useful commands