Fudge TRT transformed variable

Merged Joshua Eldon Puddefoot requested to merge ATLAS-EGamma/athena:fixTRTSmearing into master

In the ongoing optimization of HLT electronID we need to fudge a new variable. This variable is the so called transformed_e_probability_ht. There are two differences between all the other variables we fudge with EGammaVariableCorrection

  1. this variable is a variable of the track, not of the electron (so it cannot easily modified when looping on copies of electrons)
  2. what we want to fudge is not the variable as stored in the AOD, but a transformed version. This transformed version is used during the idenfication by AsgElectronLikelihoodTool

About the second point, presently (before this MR) the transformed version is computed internally by AsgElectronLikelihoodTool

The strategy in this MR:

  • AsgElectronLikelihoodTool searches for the electron decoration transformed_e_probability_ht. If it is not present (e.g. running on data, running at reconstruction, ...) it proceeds as usual: it reads eProbabilityHT from the track and it computes the transformed version and uses it (a double)
  • if EGammaVariableCorrection is called before the identification tool it applies fudging to the variables as usual, changing their values. A special case is created for transformed_e_probability_ht. If this is present in the list, the tool reads eProbabilityHT from the track and creates the tranformed version and stores that as a decoration of the electron (as a float). Then it proceeds as usual (i.e. fudging it).

@christos @jpuddefo @psommer @fnechans @jdevivi @ludovica @ahadef @gtarna

Edited by Ruggero Turra

Merge request reports