kit-cms / fitmetrecoilunc Goto Github PK

In order to setup the software, follow the commands in the checkout.sh or simply download and execute it:

wget https://raw.githubusercontent.com/KIT-CMS/FitMETRecoilUnc/master/scripts/checkout.sh
bash checkout.sh

Before creating datacards, please copy root files with input shapes into the shapes folder of this package. The root files should have the following naming scheme:

htt_mm.inputs-sm-Run${YEAR}-ML-${VARIABLE}.root

Prepare inputs for YEAR in 2016 2017 2018 and the MET component parallel to Z $\vec{p}_{T}$, called e.g. metParToZ.

The files must contain the following categories as folders:

• mm_0jet
• mm_1jet
• mm_ge2jet

Futhermore, distriutions of the following processes must be provided:

• ZL (signal process)
• ZTT
• TTT
• VVT
• TTL
• VVL
• W
• QCD

Only for the ZL process, systematic variations are introduced being 10% shifts in the scale and resolution of the hadronic recoil. These variations should have the following names:

• CMS_htt_boson_scale_met_Run${YEAR}Up
• CMS_htt_boson_scale_met_Run${YEAR}Down
• CMS_htt_boson_reso_met_Run${YEAR}Up
• CMS_htt_boson_reso_met_Run${YEAR}Down

Feel free to extend the uncertainty model, since this one is far from complete.

The first step of datacard creation is covered by the MorphingMETRecoil executable. Asuming, that datacards for all Run II years are needed, using metParToZ and puppimetParToZ as variables, the following shell command applies:

for var in metParToZ puppimetParToZ;
do
    for year in $(seq 2016 2018);
    do
        MorphingMETRecoil --era=${year} --variable=${var}
    done
done

This is done with the usual text-to-workspace method of CombineHarvester separately for each category:

combineTool.py -M T2W -i output_*/Run201*/mm*/ --parallel 10

To constrain the MET recoil uncertainties, a maximum likelihood fit is performed, with ZL as signal with a floating signal strength r in order to cover normalization effects. The recoil uncertainties are treated as nuisances. Execute the following command for the fit:

combineTool.py -M FitDiagnostics -d output_*/Run201*/*/combined.txt.cmb.root --there --parallel 10

As a final step, the total uncertainty for scale and resolution is determined for each category. The computation follows this formula: $10\% \cdot \sqrt{\text{pull}^2 + \text{constraint}^2}$. The calulcation is performed with the determine_recoiluncs.py script. To perform the computation for all years and MET definitions, execute:

for var in metParToZ puppimetParToZ;
do
    for year in 2016 2017 2018;
    do
        python $CMSSW_BASE/src/CombineHarvester/METRecoil/scripts/determine_recoiluncs.py --initial_shift 0.1 --input_dir output_${var}/Run${year}/
    done
done

This will create .json files in the output_${var}/Run${year}/ directories, which can be used further to define final recoil correction uncertainties

To check, how the fit changes the distributions, the following command can be used to obtain postift and prefit shapes:

for cf in `ls output_*metParToZ/Run201*/mm*/ -d`;
do
    PostFitShapesFromWorkspace -w ${cf}combined.txt.cmb.root -d ${cf}combined.txt.cmb  --postfit -f ${cf}fitDiagnostics.Test.root:fit_s -o ${cf}postfit_shapes.root
done