[WIP] add plotting script to produce jet energy resolution plots #129
Conversation
There was a problem hiding this comment.
I would suggest to rather look at the relative response of the matched jet pair, then doing an overall energy spread resolution measure. At very large energies, like they are present at CLIC's highest energies e.g., we have cases where the dijet exclusive jet cluster catches two out of 3 significant jets. These cases are indeed pretty rare, maybe a dphi requirement between both reconstructed jets and both MC jets dphi > 2.8 could ensure we cut out these events before filling the histograms as well and leave the code as is otherwise.
| class histPool(object): | ||
| """Helper class which contains all histograms.""" | ||
|
|
||
| cosThetaRange = [0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.925, |
There was a problem hiding this comment.
| cosThetaRange = [0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.925, | |
| cosThetaRange = [0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.925, | |
| 0.95, 0.975, 0.985, 1.025] |
in the end we always hadd the trouble with the last bin, thus i introduced another bin from 0.975 to 0.985 where i plotted the last jet energy resolution, going up to that point and then another bin up to 1.025. The last bin was left empty, as there we have jets which are partially outside of the Calorimeter range, so in the C++ root macro it would have been
const unsigned int nRegionBins(14);
float pRegionBinEdges[nRegionBins + 1] = {0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.925, 0.95, 0.975, 0.985,1.025};
| """???""" | ||
| outHistInfo = [] | ||
| histInfoTemplate = ['TH1F', 'cosThetaBin', '; reco + true neutrino energy [GeV]; Entries', | ||
| 10000, 0., 500.] |
There was a problem hiding this comment.
for CLIC the upper limit might need larger numbers, for sure larger than 1500., I even observed very high tail events up to 3000.
| for i in range(1, len(self.cosThetaRange)): | ||
| if cosTheta > self.cosThetaRange[i - 1] and cosTheta < self.cosThetaRange[i]: | ||
| binNumber = i - 1 | ||
| fillVal = vTuple.recoJets[0].E() + vTuple.totalTrueInvisible.E() |
There was a problem hiding this comment.
typically the amount of invisible energy is negligible, I think maybe calculating the response value between the matched jets might be rather the value to fill, reconstructed-jet E/ matched MC particle-jet E, and for jet 0 and jet 1, then the histogram limits would also need no energy dependent range for CLIC and FCC, we could just have a range from 0 to 2 (or if we believe things to go south up to 3).
Final results are close to each other in both cases, if we want to correct the bias of the invisible energy, maybe a split of it to both jets would be better on average.
| resolution = rms90Data['rms90'] | ||
| resolutionError = resolution / math.sqrt(rms90Data['totalCounts']) | ||
|
|
||
| self.histDict2D[histSetName]['JER'].SetBinContent(i, resolution) |
There was a problem hiding this comment.
do we want to fill the actual jet energy resolution value instead of the relative value, aka JER/jet-energy or JER/Mean. Iwould rather do the later, in case the calibration of the jet energy would be off as additional effect.
This can also be done in post processing as well though
| recoJets=getLorentzVectors(iEv.recoJetPx, iEv.recoJetPy, iEv.recoJetPz, iEv.recoJetE) | ||
| ) | ||
|
|
||
| if not checkJetPairMatching(vTuple.genJets, vTuple.recoJets, 10./180.*math.pi): |
There was a problem hiding this comment.
when checking of the jet pair match, at this point I would also save the responses of the matched pair,
resp = E_recoJet/E_genJet
as well as dphi = delta(phi_genJet,phi_recoJet) and dtheta = theta_recoJet-theta_genJet
thus we have all 3 quantities needed for filling the histograms, each jet is filled then
|
should we resume work on here? |
BEGINRELEASENOTES
ENDRELEASENOTES