Sven Menke (e5163ae0) at 27 Mar 10:10
saving event wide weight_bTagSF_off1_off2_onl_ and corresponding sy...
sorry, I was not very clear on this point. For each tool there is a separate instance. This holds for all 6 WPs we use in the analysis. But they are kept in two different lists - one for the offline WPs and one for online and conditional WPs. Hence the treatment for offline differs from the treatment for online and conditional.
I've created a JIRA here: https://its.cern.ch/jira/browse/ANALYSISTO-1391
o.k. I can check - the code runs fine for us since about 4 weeks in 21.2.271 ... do you have a log file?
o.k. yes, that would make this convoluted mess easier to read ...
same here - specifying each WP as full pair has the advantage of full configuration control in the steering file without relying on assumptions about the actual combination needed.
no, because the names of the WP are set in two different CDI files - one for online and conditional and one for offline. If needed one could configure the tool to pair any two WP with each other and then each needs to be a full pair that can be automatically parsed. Hence I would propose to keep the code as is.
probably not. These are copied from the corresponding offline b-tagging code ... I'll remove them.
no, the first is checking m_calo_WPs while the second tests m_calo_WPs_calib
o.k. then I'll do it for all 8 ... Thanks, --Sven
there are these 4 already: https://acode-browser1.usatlas.bnl.gov/lxr/source/athena/PhysicsAnalysis/TopPhys/xAOD/TopAnalysis/Root/EventSaverFlatNtuple.cxx?v=21.2#0923 https://acode-browser1.usatlas.bnl.gov/lxr/source/athena/PhysicsAnalysis/TopPhys/xAOD/TopAnalysis/Root/EventSaverFlatNtuple.cxx?v=21.2#0951 https://acode-browser1.usatlas.bnl.gov/lxr/source/athena/PhysicsAnalysis/TopPhys/xAOD/TopAnalysis/Root/EventSaverFlatNtuple.cxx?v=21.2#2317 https://acode-browser1.usatlas.bnl.gov/lxr/source/athena/PhysicsAnalysis/TopPhys/xAOD/TopAnalysis/Root/EventSaverFlatNtuple.cxx?v=21.2#2335 So should we make the 4 new ones different from the 4 existing ones?
this was following the example of the prior existing 4 loops of similar nature for named systematics. I see that you made the same suggestion for the other 3 we added. Do you want us to change those new instances only? Or should we apply it to all 8? I.e. also the 4 pre-existing ones we did not touch? And what would be wrong with const auto& name
in these cases? Thanks, --Sven
yes, currently we have two separate instances - one for offline as one for online + conditional. So, indeed, I think you are right that we can get rid of this one check of Online in the name since only the getScaleFactor method with 3 arguments is used for onl and cond WP.
it comes through getScaleFactor which checks for fillVariables
I think the question was about line 1056 in BTaggingEfficiencyTool.cxx. The only addition in that line is the check for "Online" in the name of the WP. The conditional WP that depend on any of the offline cases checked here (DL1 in 21.2, DL1, GN1 and GN2 in main) need to be excluded since they do not provide this functionality. So in my mind this change will have to be done in main like it is proposed in this MR for 21.2.
I answered 'yes' to Thomas' question, if this needs to be propageted to main.
yes, see here: https://gitlab.cern.ch/atlas/athena/-/blob/main/PhysicsAnalysis/JetTagging/JetTagPerformanceCalibration/xAODBTaggingEfficiency/Root/BTaggingEfficiencyTool.cxx?ref_type=heads#L1311 The reason is that the conditional WP have very long names including the offline WP they depend on - and if that has DL1 in the name it would trigger the DL1 treatment, which is not correct in the conditional case.
Once this is included in 21.2 we can think about how to migrate all to main. But this requires time beyond finishing the ttbb all-hadronic analysis. So currently the code is needed in 21.2 only.
The recommendations for b-tagging scale factors and efficiencies that combine offline WPs with online WPs require changes to JetTagging/JetTagPerformanceCalibration/xAODBTaggingEfficiency and 11 TopPhys packages in order to include b-tagging systematics and MC to data scale factors in 21.1-based top analyses. The proposed changes here allow to combine one offline WP with a online WP via the provided conditional (online given offline) probabilities provided by the CP-group. The offline trigger confirmation requires furthermore to combine two offline WP (from the same tagger) with the online WP. Also this functionality is provided in the MR.