Non-Gaussian effects in trigger efficiency fit

The fit of trigger efficiency with the standard error function does not work for single jet triggers at the necessary level of precision required for UL analyses. The non-Gaussian effect in the

p_\mathrm{T}^\text{HLT}

are suspected to be the reason for this. Using the integral of a Crystal-Ball function is expected to fix the problem.

Note: to investigate the tails, one can copy and adapt the code of getJetResponse to store the HLT response. Then fitJetResponse is expected to work out of the box. Alternatively, one can simply try to fit the function with the additional parameters.

To implement the fit with the Crystal-Ball modification, one can also just adapt getTriggerTurnons:

• Modify Normalisation::FitTriggerEfficiency()
• Call this function in getTriggerTurnons

added to epic &6

added To Do label

changed the description

changed the description

assigned to @aziaka

removed To Do label

added Ongoing label

mentioned in merge request !171 (merged)

added changepatch typephysics labels

In !171 (merged), we have set up a fit of a modified HLT jet response. Its integral is close to the efficiency curve, but possibly not identical (for instance, trigger 40 does not consider any jets with HLT pt above 60, etc.). At this stage, we don't know how large the differences are.

A few observations:

1. The (modified) response is less well describe than the original response. We have to see if we can play with the binning.
2. We should also check the impact of the JES variations on the determination of the efficiency curve, and see if that should be accounted for in the fit.

In this report at SMP-HAD meeting, we show how to fit the region of high efficiency with

1-N\exp{\left(-bp_\mathrm{T}\right)}

.