Fix B tag calibrations

[watson-ij]

This is a high-effort task, probably we want to involve a one person from each of the groups.

Daniel, Yechan, can you point to the documentation so we can get people up to speed?

[SangMLee]

I am pretty sure that this is the b-tagging method that we are currently trying to use

• https://twiki.cern.ch/twiki/bin/view/CMS/BTagShapeCalibration

some additional materials

• https://twiki.cern.ch/twiki/bin/view/CMS/BtagRecommendation80XReReco#Data_MC_Scale_Factors_period_dep

[watson-ij]

https://twiki.cern.ch/twiki/bin/view/CMS/BTagCalibration ?

[watson-ij]

Or maybe https://twiki.cern.ch/twiki/bin/view/CMS/BTagSFMethods also

[watson-ij]

Theres also python code in nanoAOD-tools that seems to deal with this, its really ugly though:

https://github.com/cms-nanoAOD/nanoAOD-tools/blob/master/python/postprocessing/modules/btv/btagSFProducer.py

[SangMLee]

BTag weights now come as an event weight with the nanoAOD root files in v5 of the productions
variable: btagWeight_CSVV2

[watson-ij]

What about uncertainties?

[SangMLee]

they're not in the root file
Does that mean we are unable to use the ones in the root file ?

[watson-ij]

I assume its calculated correctly, but to do systematics we will need to get the uncertainties so we probably still need to implement it anyway

[jshlee]

is this finished? can we close?

[watson-ij]

btag systematics?

[jshlee]

fyi - nanoAOD has btagging weights by default
https://github.com/cms-sw/cmssw/blob/master/PhysicsTools/NanoAOD/python/nano_cff.py#L57-L90

@quark2 can you look into this

[quark2]

I have no idea which one (the measurementType; iterativeFit or mujets, ttbar, etc; see https://twiki.cern.ch/twiki/bin/viewauth/CMS/BtagRecommendation80XReReco#AK4%20jets) to use. There is no prescription about this. I just found that CATTools guys are using 'iterativeFit', same as nanoAOD guys, but I don't know why they chose that. Jason suggested to compare them in my analysis.

[watson-ij]

Are we happy with btagging+systs as currently implemented?

[quark2]

I think okay.
FYI (or TMI), anyone who wants to use other measurementType can replace it in his/her analyser and that's all. So, everything is neat well, I think.