This TWiki provides information of the ATLAS VHbb-resolved analysis, mainly for the LPNHE-Paris ATLAS VHbb team.

Analysis workflow

This is a generic workflow of ATLAS VHbb analysis. Starting from ATLAS collisions, it is taken off as xAOD format dataset. Three types of derivations from xAOD format dataset have been used, i.e: HIGG5D1 (0-lepton filter); HIGG5D2 (1-lepton filter); HIGG2D4 (2-leptons filter). Then the calibrated xAOD datatsets (CxAODs) are prepared by running CxAOD framework. Based on CxAODs, various studies have been done, e.g: data-vs-MC comparisons, MVA studies, modelling systematic evaluation. Final results can be extracted from histograms that are produced from CxAODs with full updates of MVA, modelling systematics. WSMaker package is used to extract result, using histograms as inputs.

xAOD and derivations

xAOD datasets have been prepared centrally by ATLAS Production Group, for collision data (automatically) and for MC (analysis group submit their request). Then the analysis group launch the request of derivations. Requests of xAOD (MC only) and derivations can be done via submitting JIRA ticket. These are the latest requests from VHbb-resolved analysis:

• xAOD requests: latest xAOD sample list is available in Gitlab repository
• derivations
  • ATLDH5REQ-63
  • ATLDH5REQ-79
  • ATLDH5REQ-85
  • ATLDH5REQ-82.

CxAOD production

The VHbb analysis uses CxAOD framework to make CxAOD and histograms production. The task is taken by a task force team. There are two steps. The first step is to make DxAOD to CxAOD production by running CxAOD-Maker. The second step is to produce Histograms from CxAOD by running CxAOD-Reader. There are well-prepared instructions for running CxAOD-Maker and CxAOD-Reader.

• checkout CxAOD framework (Maker and Reader): README

This README explains how to checkout, compiling, and run test jobs for Maker and Reader. To run massive CxAOD production, it's better to launch jobs to grid. You can launch jobs from lxplus, cc-in2p3 or any other local server. For Histograms production, it's better to launch jobs to lxplus, since it reads CxAOD files and we usually download all CxAOD file in EOS that can be read from lxplus. It's also recommended to download CxAOD files to local disk with about 10 TB for one channel with systematics. So that one can run Histograms production form local sever.

Official CxAODs and Histograms

• CxAOD production: the latest version (with systematics) in /eos/atlas/atlascerngroupdisk/phys-higgs/HSG5/Run2/VH/CxAOD_r32-15/
• Histograms production: the latest version in /eos/atlas/atlascerngroupdisk/phys-higgs/HSG5/Run2/FullRunII2019/statArea/
• MVA training Trees: /eos/atlas/atlascerngroupdisk/phys-higgs/HSG5/Run2/FullRunII2019/MVA/

In case there is any question, please contact each coordinator VHbb-taskforce

Multivariate analysis

• official TWiki for VHbb MVA: link

-- KunLiu - 2019-05-06