Preparation for combination of Heavy Resonance Searches at CMS

This is a preparatory twiki designed to organise the preparation for a combination of CMS analyses searching for heavy new resonances at predicted by BSM models such as WED or HVT models.

The goal is to outline procedures and datacard requirements that allow for a smooth combination procedure once all individual analyses have passed their approval.

Interesting thoughts to the combination and BSM models to be considered were presented here (interesting slides for combination: https://indico.cern.ch/event/881563/contributions/3730910/attachments/1981776/3301585/200205_Zucchetta.pdf).

Contacts

• CMS: Daniela Schäfer
• ATLAS-CMS combination: (twiki in preparation), e-group: cms-atlas-hrf@cernNOSPAMPLEASE.ch

Datacard requirements

General

Before pre-approval:

• Ensure orthogonality to other CMS analyses with the same final state!

At the approval stage:

• Provide two sets of datacards and workspaces for the combination:
  • As used in the analysis proper
  • With the signal rate normalised to the cross section of the BSM model times the branching fraction of the considered channel. For example: rate signal = xs_BSM * BR_XtoWW * BR_SM_final_state
  • ATTENTION: for a signal scaled to the BSM cross section an additional yield uncertainty has to be added for PDF and renormalisation scales if they are not already considered
• The datacards should be provided in the following directory structure:
• CADI_line* with subfolders datacards and workspaces
• The datacards and workspaces for the combinatio should have names ending on "_combination.*"

* analyzers should make sure that the datacards follow the requirements and naming schemes listed below

Processes

• Signal process called Signal
• Background processes Background
• Data denoted by data_obs # The mass range of the signal?

Yields

• Signal yield: xs_BSM * BR_XtoVV/ll *BR_SM_final_state

Systematics

The following naming scheme should be used for systematic uncertainties that are going to be correlated between analysis. If the systematic uncertainty is particular to the single analysis and does not need to be correlated the analyzers may choose the name by themselves under consideration of the naming scheme below:

• Each nuisance should start with CMS
• If necessary add label for different data taking years e.g. CMS2016
• Add analysis specific label e.g. for an X -> VV -> 2j analysis a good label would be VV_JJ
• Custom name of the systematic uncertainty e.g. Bin1

-- DanielaSchafer - 2020-04-23