[Rivet] Question about b-tagging efficiency in Rivet

[Andy Buckley]

On 19/09/16 15:07, Adil Jueid wrote:
> Dear Prof. Andy,
>
> I hope you are doing well.
> I am using Rivet to make some distributions for single top production at
> the LHC. However, since I am interesting on b-quark observables (energy
> in proton-proton CM, b-quark pT, ..),
> I would like to know what is the efficiency of b-tagging methods that
> can be used in Rivet.
> What is the difference between the  jet.bTags() method and the method
> which cluster the b-hadrons with a jet if the deltaR separation is < 0.3.

Hi Adil,

Please email the general Rivet developer list with questions -- I've 
CC'd it on this reply.

The efficiency of b-tagging... well, it's 100% in Rivet, by definition. 
We don't apply any detector reconstruction/ID efficiency or fake rates 
by default, so we return 100% of whatever has been deemed to be a truth tag.

Different experiments may use different definitions of what a "true 
b-jet" is, though: maybe there's a threshold on the b-hadron pT for some 
papers. Or until recently, the tagging was calibrated in a more 
MC-dependent way, using generator records of (borderline-physical) 
b-quarks rather than some choice of b-hadron.

The jet.bTagged() and jet.bTags()methods can take a Cut argument, to 
restrict the b-hadrons which will be considered as taggable. We could 
extend this to also allow more general filtering such as dR w.r.t. the 
jet axis, but that's not *currently* supported as a "one-liner". The 
unfiltered tags list is all final b-hadrons ghost-associated into the 
jet, without any pT cut, so if you require a pT cut and/or your jet 
radius is > 0.3 then bTags() or bTagged() without an argument will be a 
bit too inclusive. If it would be useful, we can try to include the more 
general filtering approach for the next release.

We also now have projections to implement detector effects, primarily 
for use in BSM analyses. The SmearedJets projection can apply b-tagging 
efficiency and fake rates, but this is relative to the baseline defined 
by the truth definition.

Andy

-- 
Dr Andy Buckley, Lecturer / Royal Society University Research Fellow
Particle Physics Expt Group, University of Glasgow