Dropping large MC truth collections from SUSY3, 5, 11, and apply thinning to taus
Hello,
This MR is implementing the changes described in ATLSUSYDPD-2048, in order to reduce the DAOD size for the upcoming reprocessing (ATLSUSYDPD-2065, ATLSUSYDPD-2066, ATLSUSYDPD-2067).
SUSY3 changes:
- only consider RNN taus in the skimming, no more (RNN OR BDT), as BDT tau ID is deprecated
- move tau ID skimming from loose WP to very loose WP, to have a chance to use the FTTF tool for fake tau background
- apply thinning to taus, keep only taus with pt > 15 GeV and only keep tau/ID tracks that are classified as charged tau track
- drop large MC truth collections not used in analysis: TruthParticles, TruthEvents, TruthVertices
- use TruthLinkRepointTool, such that reco objects are truth-matched against TRUTH3 collections instead of TruthParticles
SUSY5 changes:
- apply thinning to taus, keep only taus with pt > 15 GeV and only keep tau/ID tracks that are classified as charged tau track
- drop TruthParticles, TruthEvents
- TruthVertices was kept because there seems to be more interest in vertices in SUSY5
- use TruthLinkRepointTool
SUSY11 changes:
- apply thinning to taus, keep only taus with pt > 15 GeV and only keep tau/ID tracks that are classified as charged tau track
- drop TruthParticles (with custom truth thinning), TruthEvents, TruthVertices
- use TruthLinkRepointTool
Minor adjustments were made to "ExtraVariables".
Size reduction for DAODs:
- SUSY3 data: -5.3%
- SUSY5 data: -8.7%
- SUSY11 data: -11.3%
- SUSY3 MC: -25%
- SUSY5 MC: -6.6%
- SUSY11 MC: -13.2% (concerns few samples, SUSY11 is mostly used for data)
I set the "urgent" flag because we'd like to have the MR in quickly, so a cache can be built, and the reprocessing launched hopefully on Friday.
Cheers, Bertrand