Differences between FTK and offline track reconstruction
Compare FTK tracks to offline tracks
Look at properties such as the track transverse momentum and pseudorapidity. Are there significant differences? What do they mean, and are they consistent with what we would expect about how the tracks were reconstructed?
One could also investigate the behaviour of offline and FTK tracks for the different calorimeter components (named EM1, EM2, ... , HAD4 in samples). Therefore one first has to match the FTK and offline tracks with a delta R cut. The value could be chosen to delta R < 0.1 (0.2)
Compare charged PFlow objects built with offline tracks and FTK.
As for the original tracks, how do the tracks selected for use in PFlow change? What might this imply for the calorimeter energy subtraction?
It might be interesting at some point to also compare the matching of FTK and offline tracks to the event primary vertices -- this information is not (yet) in the Ntuples, and would have to be added.
Impact of using the FTK tracks on particle flow reconstruction
Assess how the PFlow objects change when the FTK tracks are used
Some changes are expected in the charged PFOs, so we should expect to match a different cluster population for the subtraction. How are features of the neutral PFO container modified? E.g. the number of neutral PFOs or their total/mean transverse momenta.
Are these changes consistent with what was found when examining the charged PFOs? What can we infer, and how might this affect the jet performance?
Assess how the PFlow jets change when the FTK tracks are used
We can study the pileup-dependence by matching PF jets to truth jets with pT above ~4 GeV. If we do a worse job with the subtraction, we might expect an increased rate of pure pileup jets.
More importantly, we would want to assess the resolution and response differences, where response is the reconstructed jet pT divided by the truth jet pT, while the resolution is given by the width of this distribution. Ordinarily we'd prefer to do this with fully calibrated jets, but the calibration is not trivial to derive, and therefore to begin with we can probably manage by just matching truth jets to the PF jets and looking at the ratios of the pTs.
Does this matter more for low-pT jets or high-pT jets? Compare with some relevant trigger jet pT thresholds.
Calibrate Jets with numerical inversion. Bets is to find a fit to the mean values over pt-reco and apply this correction.
After Jet calibration check for closure, look at dependency of eta-> maybe apply eta calibration. This could be done in a 3-D histogram bin-by-bin, or the same calibration method as for pT.
Assess effect of pile-up
Try to find a discrimination between high-pileup events and low-pileup events and look again at the pT response curves. --> Found discrimination with Nr. of Vertices > 15 (high pileup) and < 15 (low pileup)
Reproduce the efficiency plot (in Treereader script 'Cutflow') for several pT slices -> Normalize it!
Assess performance of different triggers when using FTK and/or simple extrapolation
Trigger efficiency plot for offline 6th jet pt (multi-jet trigger), and MET (MET-trigger).
Plot trigger rate over (online) threshold (min-bias samples needed?).
Plot rate over efficiency for several thresholds.
TODO List:
Do calibration more finely in eta, look especially at EMTopo container
Trigger efficiency plots: plot efficiency for a fixed rate -> means: find out threshold of PFlow container which matches the HLT one
look at npileup jets for different pT thresholds. plot jet multiplicities
look at charged and neutral pflow container
bump in response distribution in EMTopo container due to reconstrcuted electrons?
extend jet pt range higher then 100 GeV with greater bins
matching jet efficiency plot. Implement in code
plot calib factor over eta
_
Resolution plots with offl. PF (PFOs, when in ntuples) with different track pt threshholds
make MET figure (include FTK distribution here!)
Correlation of nVertices and inst lumi
5j45 trigger
assess properties of HLT events, which show decreasing efficiency at MET trigger at high MET
put rate in legend of efficiency plot with same efficiency at 100 MET offline
under chaged PFO pt distribution also other ratios
_
Question 1: Some ATLAS text in plots?
_
Last Studies:
Understand behaviour of cPFOs and nPFOs, match PFOs with Tracks?