This repository contains the framework for the measurement of matching efficiencies, trigger turn-on curves, and scalings for the assessment of the physics performance of the CMS Phase-2 L1 Menu.
Added cache and plotting configs for producing b/mis tag plots
Cache config picks up both seeded cone jet collections from ntuples
The "extended" seeded cone jets (those produced with extended tracks) are the ones with the b jet NN defined
The plotting config produces "standard" matching plots for the two seeded cone jet collections just to check the performance of the two are similar (before any b tagging).
The plotting config also produces b/mis tag efficiency plots vs pt and eta for a working point (cut on b jet NN) that gives a mistag rate of ~10% in a ttbar+200PU sample
Added a script plotBTagEfficiency.py and config bJetEff.yaml to overlay the b/mis tag efficiencies on the same plot - example below.
The plots above include the jet reconstruction efficiency (i.e. efficiency for a gen b jet to be reconstructed as a seeded cone jet and pass the NN cut). It wasn't clear to me how to produce efficiency plots for just the NN cut with these tools (i.e. the denominator are generator jets matched to seeded cone jets). I think this is essentially what is done when plotting the turn on curves, but with a cut on the L1 pt instead of the b jet NN. I may have hacked the tools to produce this plot (not in this PR), but I hope there's a better/official way of doing this.
plotBTagEfficiency.py
and configbJetEff.yaml
to overlay the b/mis tag efficiencies on the same plot - example below.The plots above include the jet reconstruction efficiency (i.e. efficiency for a gen b jet to be reconstructed as a seeded cone jet and pass the NN cut). It wasn't clear to me how to produce efficiency plots for just the NN cut with these tools (i.e. the denominator are generator jets matched to seeded cone jets). I think this is essentially what is done when plotting the turn on curves, but with a cut on the L1 pt instead of the b jet NN. I may have hacked the tools to produce this plot (not in this PR), but I hope there's a better/official way of doing this.
@artlbv