def UnfoldFF(rawFF, detResponse, tag):
padFF.Clear()
padFF.Divide(2)
# Detector response
ResponseNorm(detResponse)
padFF.cd(1)
detResponse.Draw('COLZ')
response = RooUnfoldResponse(rawFF, rawFF, detResponse)
bayes = RooUnfoldBayes(response, rawFF)
# Unfolding - Bayesian
ana_util.COLOR = SelectColor(0)
ana_util.MARKER = SelectMarker(0)
padFF.cd(2)
# Legend
lgd = TLegend(0.12, 0.6, 0.3, 0.85)
lgd.SetName('lgd' + tag)
lgd.SetBorderSize(0)
lgd.SetFillColor(0)
# Z measured
rawFF.SetBins(10, 0, 1.0)
DrawFF(rawFF, 'hRaw_' + tag)
lgd.AddEntry(rawFF, 'Raw')
for nIter in range(4, 5):
bayes.SetIterations(nIter)
hist = DrawFF(bayes.Hreco(0), 'hBayes' + repr(nIter) + '_' + tag)
for ix in range(1, hist.GetNbinsX()):
hist.SetBinError(ix, rawFF.GetBinError(ix))
lgd.AddEntry(hist, 'Bayes (N=%d)' % nIter)
lgd.Draw('same')
padFF.Print(printFile, 'Title:' + tag)
padFF.Write('c' + tag)
def unfold_real_data(self):
logging.info('Unfolding real data')
regularisation = int(self.params['regularisation'].m)
unfold_bg = self.params['unfold_bg'].value
unfold_eff = self.params['unfold_eff'].value
raw_data_0 = self._data['nuall']
if regularisation == 0:
raise AssertionError('Regularisation is set to 0')
# Get the inversed efficiency histogram
if not unfold_eff:
inv_eff = self.get_inv_eff()
# Load response object from disk cache
response = self.create_response()
# Background Subtraction
if unfold_bg:
raw_data_1 = raw_data_0
else:
bg_hist = self.get_bg_hist()
raw_data_1 = raw_data_0 - bg_hist
r_flat = roounfold._flatten_to_1d(raw_data_1)
r_th1d = convert_to_th1d(r_flat, errors=True)
# Unfold
unfold = RooUnfoldBayes(response, r_th1d, regularisation)
unfold.SetVerbose(0)
unfolded_flat = unfold.Hreco(1)
unfold_map = unflatten_thist(in_th1d=unfolded_flat,
binning=self.true_binning,
name=self.output_str,
errors=True)
# Efficiency correction
if not unfold_eff:
unfold_map *= inv_eff
del r_th1d
del unfold
logging.info('Unfolded reco sum {0}'.format(
np.sum(unp.nominal_values(unfold_map.hist))))
return unfold_map
def unfolderFactory(self, response, hMeas, BasisSpline_m0=0):
if "Bayes" in self.optunf:
# Bayes unfoldung until chi^2 cut (max 10):
unfold = RooUnfoldBayes(response, hMeas, 10, False, True)
elif "SVD" in self.optunf:
# SVD unfolding with free regularisation:
unfold = RooUnfoldSvd(response, hMeas)
elif "TUnfold" in self.optunf:
# TUnfold with fixed regularisation tau=0.002
# unfold= RooUnfoldTUnfold( response, hMeas, 0.002 )
unfold = RooUnfoldTUnfold(response, hMeas)
elif "Invert" in self.optunf:
unfold = RooUnfoldInvert(response, hMeas)
elif "Reverse" in self.optunf:
# Use equivalent Bayes with 1 iteration:
unfold = RooUnfoldBayes(response, hMeas, 1)
elif "BasisSplines" in self.optunf:
# BasisSplines with automatic tau determination
unfold = RooUnfoldBasisSplines(response, hMeas, self.nrebin, 0.0,
BasisSpline_m0, 2)
return unfold
def setup_unfolding ( self, data ):
self.data = data
if not self.unfoldObject:
if not self.unfoldResponse:
self.unfoldResponse = self._makeUnfoldResponse()
if self.method == 'RooUnfoldBayes':
self.unfoldObject = RooUnfoldBayes ( self.unfoldResponse, self.data, self.Bayes_n_repeat )
elif self.method == 'RooUnfoldBinByBin':
self.unfoldObject = RooUnfoldBinByBin ( self.unfoldResponse, self.data )
elif self.method == 'RooUnfoldInvert':
self.unfoldObject = RooUnfoldInvert ( self.unfoldResponse, self.data )
elif self.method == 'RooUnfoldSvd':
if self.k_value > 0:
self.unfoldObject = RooUnfoldSvd( self.unfoldResponse, self.data, self.k_value, self.n_toy )
else:
if self.tau >= 0:
self.unfoldObject = RooUnfoldSvd( self.unfoldResponse, self.data, self.tau, self.n_toy )
elif self.method == 'RooUnfoldTUnfold':
self.unfoldObject = RooUnfoldTUnfold ( self.unfoldResponse, data )
if self.tau >= 0:
self.unfoldObject.FixTau( self.tau )
elif self.method == 'TSVDUnfold':
new_data = Hist( list( self.data.xedges() ), type = 'D' )
new_data.Add( self.data )
new_measured = Hist( list( self.measured.xedges() ), type = 'D' )
new_measured.Add( self.measured )
new_truth = Hist( list( self.truth.xedges() ), type = 'D' )
new_truth.Add( self.truth )
if self.fakes:
new_fakes = Hist( list ( self.fakes.xedges() ), type = 'D' )
new_fakes.Add ( self.fakes )
new_measured = new_measured - new_fakes
new_response = Hist2D( list( self.response.xedges() ), list( self.response.yedges() ), type = 'D' )
new_response.Add( self.response )
# replace global objects with new ones
self.data = new_data
self.measured = new_measured
self.truth = new_truth
self.response = new_response
self.unfoldObject = TSVDUnfold( self.data, self.measured, self.truth, self.response )
def main():
if len(sys.argv) < 3:
print("Usage: ToyMC [numberEvents] [randomSeed]")
return
numberEvents = int(sys.argv[1])
seed = int(sys.argv[2])
print(
"==================================== TRAIN ===================================="
)
f = root_open(
"legotrain_350_20161117-2106_LHCb4_fix_CF_pPb_MC_ptHardMerged.root", "read"
)
hJetPt = f.Get("AliJJetJtTask/AliJJetJtHistManager/JetPt/JetPtNFin{:02d}".format(2))
hZ = f.Get("AliJJetJtTask/AliJJetJtHistManager/Z/ZNFin{:02d}".format(2))
FillFakes = False
dummy_variable = True
weight = True
NBINS = 50
LimL = 0.1
LimH = 500
logBW = (TMath.Log(LimH) - TMath.Log(LimL)) / NBINS
LogBinsX = [LimL * math.exp(ij * logBW) for ij in range(0, NBINS + 1)]
hJetPtMeas = Hist(LogBinsX)
hJetPtTrue = Hist(LogBinsX)
myRandom = TRandom3(seed)
fEff = TF1("fEff", "1-0.5*exp(-x)")
jetBinBorders = [5, 10, 20, 30, 40, 60, 80, 100, 150, 500]
hJetPtMeasCoarse = Hist(jetBinBorders)
hJetPtTrueCoarse = Hist(jetBinBorders)
NBINSJt = 64
low = 0.01
high = 10
BinW = (TMath.Log(high) - TMath.Log(low)) / NBINSJt
LogBinsJt = [low * math.exp(i * BinW) for i in range(NBINSJt + 1)]
hJtTrue = Hist(LogBinsJt)
hJtMeas = Hist(LogBinsJt)
hJtFake = Hist(LogBinsJt)
LogBinsPt = jetBinBorders
jetPtBins = [(a, b) for a, b in zip(jetBinBorders, jetBinBorders[1:])]
hJtTrue2D = Hist2D(LogBinsJt, LogBinsPt)
hJtMeas2D = Hist2D(LogBinsJt, LogBinsPt)
hJtFake2D = Hist2D(LogBinsJt, LogBinsPt)
hJtMeasBin = [Hist(LogBinsJt) for i in jetBinBorders]
hJtTrueBin = [Hist(LogBinsJt) for i in jetBinBorders]
response = RooUnfoldResponse(hJtMeas, hJtTrue)
response2D = RooUnfoldResponse(hJtMeas2D, hJtTrue2D)
responseBin = [RooUnfoldResponse(hJtMeas, hJtTrue) for i in jetBinBorders]
responseJetPt = RooUnfoldResponse(hJetPtMeas, hJetPtTrue)
responseJetPtCoarse = RooUnfoldResponse(hJetPtMeasCoarse, hJetPtTrueCoarse)
# Histogram index is jet pT index, Bin 0 is 5-10 GeV
# Histogram X axis is observed jT, Bin 0 is underflow
# Histogram Y axis is observed jet Pt, Bin 0 is underflow
# Histogram Z axis is True jT, Bin 0 is underflow
responses = [Hist3D(LogBinsJt, LogBinsPt, LogBinsJt) for i in jetPtBins]
misses = Hist2D(LogBinsJt, LogBinsPt)
fakes2D = Hist2D(LogBinsJt, LogBinsPt)
outFile = TFile("tuple.root", "recreate")
responseTuple = TNtuple(
"responseTuple", "responseTuple", "jtObs:ptObs:jtTrue:ptTrue"
)
hMultiTrue = Hist(50, 0, 50)
hMultiMeas = Hist(50, 0, 50)
hZMeas = Hist(50, 0, 1)
hZTrue = Hist(50, 0, 1)
hZFake = Hist(50, 0, 1)
responseMatrix = Hist2D(LogBinsJt, LogBinsJt)
numberJets = 0
numberFakes = 0
numberJetsMeasBin = [0 for i in jetBinBorders]
numberJetsTrueBin = [0 for i in jetBinBorders]
numberFakesBin = [0 for i in jetBinBorders]
ieout = numberEvents / 10
if ieout > 10000:
ieout = 10000
fakeRate = 1
start_time = datetime.now()
print("Processing Training Events")
for ievt in range(numberEvents):
tracksTrue = []
tracksMeas = [0 for x in range(100)]
if ievt % ieout == 0 and ievt > 0:
time_elapsed = datetime.now() - start_time
time_left = timedelta(
seconds=time_elapsed.total_seconds()
* 1.0
* (numberEvents - ievt)
/ ievt
)
print(
"Event {} [{:.2f}%] Time Elapsed: {} ETA: {}".format(
ievt,
100.0 * ievt / numberEvents,
fmtDelta(time_elapsed),
fmtDelta(time_left),
)
)
jetTrue = TVector3(0, 0, 0)
jetMeas = TVector3(0, 0, 0)
jetPt = hJetPt.GetRandom()
remainder = jetPt
if jetPt < 5:
continue
nt = 0
nt_meas = 0
while remainder > 0:
trackPt = hZ.GetRandom() * jetPt
if trackPt < remainder:
track = TVector3()
remainder = remainder - trackPt
else:
trackPt = remainder
remainder = -1
if trackPt > 0.15:
track.SetPtEtaPhi(
trackPt, myRandom.Gaus(0, 0.1), myRandom.Gaus(math.pi, 0.2)
)
tracksTrue.append(track)
jetTrue += track
if fEff.Eval(trackPt) > myRandom.Uniform(0, 1):
tracksMeas[nt] = 1
jetMeas += track
nt_meas += 1
else:
tracksMeas[nt] = 0
nt += 1
fakes = []
for it in range(fakeRate * 100):
if myRandom.Uniform(0, 1) > 0.99:
fake = TVector3()
fake.SetPtEtaPhi(
myRandom.Uniform(0.15, 1),
myRandom.Gaus(0, 0.1),
myRandom.Gaus(math.pi, 0.2),
)
fakes.append(fake)
jetMeas += fake
hJetPtMeas.Fill(jetMeas.Pt())
hJetPtTrue.Fill(jetTrue.Pt())
responseJetPt.Fill(jetMeas.Pt(), jetTrue.Pt())
responseJetPtCoarse.Fill(jetMeas.Pt(), jetTrue.Pt())
hMultiTrue.Fill(nt)
hMultiMeas.Fill(nt_meas)
ij_meas = GetBin(jetBinBorders, jetMeas.Pt())
ij_true = GetBin(jetBinBorders, jetTrue.Pt())
if nt < 5 or nt_meas < 5:
continue
numberJets += 1
if ij_meas >= 0:
numberJetsMeasBin[ij_meas] += 1
hJetPtMeasCoarse.Fill(jetMeas.Pt())
if ij_true >= 0:
numberJetsTrueBin[ij_true] += 1
hJetPtTrueCoarse.Fill(jetTrue.Pt())
for track, it in zip(tracksTrue, range(100)):
zTrue = (track * jetTrue.Unit()) / jetTrue.Mag()
jtTrue = (track - scaleJet(jetTrue, zTrue)).Mag()
hZTrue.Fill(zTrue)
if ij_true >= 0:
if weight:
hJtTrue.Fill(jtTrue, 1.0 / jtTrue)
hJtTrueBin[ij_true].Fill(jtTrue, 1.0 / jtTrue)
hJtTrue2D.Fill(jtTrue, jetTrue.Pt(), 1.0 / jtTrue)
else:
hJtTrue.Fill(jtTrue)
hJtTrueBin[ij_true].Fill(jtTrue)
hJtTrue2D.Fill(jtTrue, jetTrue.Pt())
if ij_meas >= 0:
if tracksMeas[it] == 1:
zMeas = (track * jetMeas.Unit()) / jetMeas.Mag()
jtMeas = (track - scaleJet(jetMeas, zMeas)).Mag()
hZMeas.Fill(zMeas)
if weight:
hJtMeasBin[ij_meas].Fill(jtMeas, 1.0 / jtMeas)
hJtMeas.Fill(jtMeas, 1.0 / jtMeas)
hJtMeas2D.Fill(jtMeas, jetMeas.Pt(), 1.0 / jtMeas)
else:
hJtMeas.Fill(jtMeas)
hJtMeasBin[ij_meas].Fill(jtMeas)
hJtMeas2D.Fill(jtMeas, jetMeas.Pt())
response.Fill(jtMeas, jtTrue)
responseBin[ij_true].Fill(jtMeas, jtTrue)
response2D.Fill(jtMeas, jetMeas.Pt(), jtTrue, jetTrue.Pt())
responseMatrix.Fill(jtMeas, jtTrue)
responses[ij_true].Fill(jtMeas, jetMeas.Pt(), jtTrue)
responseTuple.Fill(jtMeas, jetMeas.Pt(), jtTrue, jetTrue.Pt())
else:
response.Miss(jtTrue)
responseBin[ij_true].Miss(jtTrue)
response2D.Miss(jtTrue, jetTrue.Pt())
misses.Fill(jtTrue, jetTrue.Pt())
responseTuple.Fill(-1, -1, jtTrue, jetTrue.Pt())
if ij_meas >= 0:
for fake in fakes:
zFake = (fake * jetMeas.Unit()) / jetMeas.Mag()
jtFake = (fake - scaleJet(jetMeas, zFake)).Mag()
hZMeas.Fill(zFake)
hZFake.Fill(zFake)
if weight:
hJtMeas.Fill(jtFake, 1.0 / jtFake)
hJtMeasBin[ij_meas].Fill(jtFake, 1.0 / jtFake)
hJtMeas2D.Fill(jtFake, jetMeas.Pt(), 1.0 / jtFake)
hJtFake2D.Fill(jtFake, jetMeas.Pt(), 1.0 / jtFake)
hJtFake.Fill(jtFake, 1.0 / jtFake)
else:
hJtMeas.Fill(jtFake)
hJtMeasBin[ij_meas].Fill(jtFake)
hJtMeas2D.Fill(jtFake, jetMeas.Pt())
hJtFake2D.Fill(jtFake, jetMeas.Pt())
hJtFake.Fill(jtFake)
if FillFakes:
response.Fake(jtFake)
responseBin[ij_true].Fake(jtFake)
response2D.Fake(jtFake, jetMeas.Pt())
fakes2D.Fill(jtFake, jetMeas.Pt())
responseTuple.Fill(jtFake, jetMeas.Pt(), -1, -1)
numberFakes += 1
numberFakesBin[ij_true] += 1
response2Dtest = make2Dresponse(
responses, jetPtBins, hJtMeas2D, hJtTrue2D, misses=misses, fakes=fakes2D
)
if dummy_variable:
hJetPtMeas.Reset()
hJetPtTrue.Reset()
hMultiTrue.Reset()
hMultiMeas.Reset()
hJetPtMeasCoarse.Reset()
hJetPtTrueCoarse.Reset()
hZTrue.Reset()
hZMeas.Reset()
hJtTrue.Reset()
hJtTrue2D.Reset()
hJtMeas.Reset()
hJtMeas2D.Reset()
hJtFake.Reset()
hJtFake2D.Reset()
for h, h2 in zip(hJtTrueBin, hJtMeasBin):
h.Reset()
h2.Reset()
numberJetsMeasBin = [0 for i in jetBinBorders]
numberJetsTrueBin = [0 for i in jetBinBorders]
numberJets = 0
print("Create testing data")
start_time = datetime.now()
numberEvents = numberEvents / 2
for ievt in range(numberEvents):
tracksTrue = []
tracksMeas = [0 for x in range(100)]
if ievt % ieout == 0 and ievt > 0:
time_elapsed = datetime.now() - start_time
time_left = timedelta(
seconds=time_elapsed.total_seconds()
* 1.0
* (numberEvents - ievt)
/ ievt
)
print(
"Event {} [{:.2f}%] Time Elapsed: {} ETA: {}".format(
ievt,
100.0 * ievt / numberEvents,
fmtDelta(time_elapsed),
fmtDelta(time_left),
)
)
jetTrue = TVector3(0, 0, 0)
jetMeas = TVector3(0, 0, 0)
jetPt = hJetPt.GetRandom()
remainder = jetPt
if jetPt < 5:
continue
nt = 0
nt_meas = 0
while remainder > 0:
trackPt = hZ.GetRandom() * jetPt
if trackPt < remainder:
track = TVector3()
remainder = remainder - trackPt
else:
trackPt = remainder
remainder = -1
if trackPt > 0.15:
track.SetPtEtaPhi(
trackPt, myRandom.Gaus(0, 0.1), myRandom.Gaus(math.pi, 0.2)
)
tracksTrue.append(track)
jetTrue += track
if fEff.Eval(trackPt) > myRandom.Uniform(0, 1):
tracksMeas[nt] = 1
jetMeas += track
nt_meas += 1
else:
tracksMeas[nt] = 0
nt += 1
fakes = []
for it in range(fakeRate * 100):
if myRandom.Uniform(0, 1) > 0.99:
fake = TVector3()
fake.SetPtEtaPhi(
myRandom.Uniform(0.15, 1),
myRandom.Gaus(0, 0.1),
myRandom.Gaus(math.pi, 0.2),
)
fakes.append(fake)
jetMeas += fake
hJetPtMeas.Fill(jetMeas.Pt())
hJetPtTrue.Fill(jetTrue.Pt())
hMultiTrue.Fill(nt)
hMultiMeas.Fill(nt_meas)
ij_meas = GetBin(jetBinBorders, jetMeas.Pt())
ij_true = GetBin(jetBinBorders, jetTrue.Pt())
if nt < 5 or nt_meas < 5:
continue
numberJets += 1
if ij_meas >= 0:
numberJetsMeasBin[ij_meas] += 1
hJetPtMeasCoarse.Fill(jetMeas.Pt())
if ij_true >= 0:
numberJetsTrueBin[ij_true] += 1
hJetPtTrueCoarse.Fill(jetTrue.Pt())
for track, it in zip(tracksTrue, range(100)):
zTrue = (track * jetTrue.Unit()) / jetTrue.Mag()
jtTrue = (track - scaleJet(jetTrue, zTrue)).Mag()
hZTrue.Fill(zTrue)
if ij_true >= 0:
if weight:
hJtTrue.Fill(jtTrue, 1.0 / jtTrue)
hJtTrueBin[ij_true].Fill(jtTrue, 1.0 / jtTrue)
hJtTrue2D.Fill(jtTrue, jetTrue.Pt(), 1.0 / jtTrue)
else:
hJtTrue.Fill(jtTrue)
hJtTrueBin[ij_true].Fill(jtTrue)
hJtTrue2D.Fill(jtTrue, jetTrue.Pt())
if ij_meas >= 0:
if tracksMeas[it] == 1:
zMeas = (track * jetMeas.Unit()) / jetMeas.Mag()
jtMeas = (track - scaleJet(jetMeas, zMeas)).Mag()
hZMeas.Fill(zMeas)
if weight:
hJtMeasBin[ij_meas].Fill(jtMeas, 1.0 / jtMeas)
hJtMeas.Fill(jtMeas, 1.0 / jtMeas)
hJtMeas2D.Fill(jtMeas, jetMeas.Pt(), 1.0 / jtMeas)
else:
hJtMeas.Fill(jtMeas)
hJtMeasBin[ij_meas].Fill(jtMeas)
hJtMeas2D.Fill(jtMeas, jetMeas.Pt())
if ij_meas >= 0:
for fake in fakes:
zFake = (fake * jetMeas.Unit()) / jetMeas.Mag()
jtFake = (fake - scaleJet(jetMeas, zFake)).Mag()
hZMeas.Fill(zFake)
hZFake.Fill(zFake)
if weight:
hJtMeas.Fill(jtFake, 1.0 / jtFake)
hJtMeasBin[ij_meas].Fill(jtFake, 1.0 / jtFake)
hJtMeas2D.Fill(jtFake, jetMeas.Pt(), 1.0 / jtFake)
hJtFake2D.Fill(jtFake, jetMeas.Pt(), 1.0 / jtFake)
hJtFake.Fill(jtFake, 1.0 / jtFake)
else:
hJtMeas.Fill(jtFake)
hJtMeasBin[ij_meas].Fill(jtFake)
hJtMeas2D.Fill(jtFake, jetMeas.Pt())
hJtFake2D.Fill(jtFake, jetMeas.Pt())
hJtFake.Fill(jtFake)
time_elapsed = datetime.now() - start_time
print(
"Event {} [{:.2f}%] Time Elapsed: {}".format(
numberEvents, 100.0, fmtDelta(time_elapsed)
)
)
if not FillFakes:
hJtMeas.Add(hJtFake, -1)
hJtMeas2D.Add(hJtFake2D, -1)
responseTuple.Print()
outFile.Write()
# printTuple(responseTuple)
hJtMeasProjBin = [
makeHist(hJtMeas2D.ProjectionX("histMeas{}".format(i), i, i), bins=LogBinsJt)
for i in range(1, len(jetBinBorders))
]
hJtMeasProj = makeHist(hJtMeas2D.ProjectionX("histMeas"), bins=LogBinsJt)
hJtTrueProjBin = [
makeHist(hJtTrue2D.ProjectionX("histTrue{}".format(i), i, i), bins=LogBinsJt)
for i in range(1, len(jetBinBorders))
]
hJtTrueProj = makeHist(hJtTrue2D.ProjectionX("histTrue"), bins=LogBinsJt)
hJtFakeProjBin = [
makeHist(hJtFake2D.ProjectionX("histFake{}".format(i), i, i), bins=LogBinsJt)
for i in range(1, len(jetBinBorders))
]
if not FillFakes:
for h, h2 in zip(hJtMeasBin, hJtFakeProjBin):
h.Add(h2, -1)
for h in (
hJtMeasProj,
hJtTrueProj,
hJtMeas,
hJtTrue,
hJtFake,
hZFake,
hZMeas,
hZTrue,
):
h.Scale(1.0 / numberJets, "width")
for meas, true, n_meas, n_true in zip(
hJtMeasBin, hJtTrueBin, numberJetsMeasBin, numberJetsTrueBin
):
if n_meas > 0:
meas.Scale(1.0 / n_meas, "width")
if n_true > 0:
true.Scale(1.0 / n_true, "width")
numberJetsMeasFromHist = [
hJetPtMeasCoarse.GetBinContent(i)
for i in range(1, hJetPtMeasCoarse.GetNbinsX() + 1)
]
numberJetsTrueFromHist = [
hJetPtTrueCoarse.GetBinContent(i)
for i in range(1, hJetPtTrueCoarse.GetNbinsX() + 1)
]
print("Total number of jets: {}".format(numberJets))
print("Total number of fakes: {}".format(numberFakes))
print("Measured jets by bin")
print(numberJetsMeasBin)
print(numberJetsMeasFromHist)
print("True jets by bin")
print(numberJetsTrueBin)
print(numberJetsTrueFromHist)
hRecoJetPtCoarse = unfoldJetPt(hJetPtMeasCoarse, responseJetPtCoarse, jetBinBorders)
numberJetsFromReco = [
hRecoJetPtCoarse.GetBinContent(i)
for i in range(1, hRecoJetPtCoarse.GetNbinsX())
]
print("Unfolded jet numbers by bin:")
print(numberJetsFromReco)
print("Fakes by bin")
print(numberFakesBin)
print(
"==================================== UNFOLD ==================================="
)
unfold = RooUnfoldBayes(response, hJtMeas, 4) # OR
unfoldSVD = RooUnfoldSvd(response, hJtMeas, 20) # OR
unfold2D = RooUnfoldBayes(response2D, hJtMeas2D, 4)
for u in (unfold, unfoldSVD, unfold2D):
u.SetVerbose(0)
# response2Dtest = makeResponseFromTuple(responseTuple,hJtMeas2D,hJtTrue2D)
unfold2Dtest = RooUnfoldBayes(response2Dtest, hJtMeas2D, 4)
unfoldBin = [
RooUnfoldBayes(responseBin[i], hJtMeasBin[i]) for i in range(len(jetBinBorders))
]
for u in unfoldBin:
u.SetVerbose(0)
hRecoBayes = makeHist(unfold.Hreco(), bins=LogBinsJt)
hRecoSVD = makeHist(unfoldSVD.Hreco(), bins=LogBinsJt)
hRecoBin = [
makeHist(unfoldBin[i].Hreco(), bins=LogBinsJt)
for i in range(len(jetBinBorders))
]
hReco2D = make2DHist(unfold2D.Hreco(), xbins=LogBinsJt, ybins=LogBinsPt)
hReco2Dtest = make2DHist(unfold2Dtest.Hreco(), xbins=LogBinsJt, ybins=LogBinsPt)
hRecoJetPt = unfoldJetPt(hJetPtMeas, responseJetPt, LogBinsX)
hReco2DProjBin = [
makeHist(hReco2D.ProjectionX("histReco{}".format(i), i, i), bins=LogBinsJt)
for i in range(1, len(jetBinBorders))
]
hReco2DTestProjBin = [
makeHist(
hReco2Dtest.ProjectionX("histRecoTest{}".format(i), i, i), bins=LogBinsJt
)
for i in range(1, len(jetBinBorders))
]
hReco2DProj = makeHist(hReco2D.ProjectionX("histReco"), bins=LogBinsJt)
hReco2DProj.Scale(1.0 / numberJets, "width")
for h, h2, n in zip(hReco2DProjBin, hReco2DTestProjBin, numberJetsFromReco):
if n > 0:
h.Scale(1.0 / n, "width")
h2.Scale(1.0 / n, "width")
# unfold.PrintTable (cout, hJtTrue)
for h, h2, nj in zip(hJtMeasProjBin, hJtFakeProjBin, numberJetsMeasBin):
if nj > 0:
h.Scale(1.0 / nj, "width")
h2.Scale(1.0 / nj, "width")
# else:
# print("nj is 0 for {}".format(h.GetName()))
for h, nj in zip(hJtTrueProjBin, numberJetsTrueBin):
if nj > 0:
h.Scale(1.0 / nj, "width")
# draw8grid(hJtMeasBin[1:],hJtTrueBin[1:],jetPtBins[1:],xlog = True,ylog = True,name="newfile.pdf",proj = hJtMeasProjBin[2:], unf2d = hReco2DProjBin[2:], unf=hRecoBin[1:])
if numberEvents > 1000:
if numberEvents > 1000000:
filename = "ToyMC_{}M_events.pdf".format(numberEvents / 1000000)
else:
filename = "ToyMC_{}k_events.pdf".format(numberEvents / 1000)
else:
filename = "ToyMC_{}_events.pdf".format(numberEvents)
draw8gridcomparison(
hJtMeasBin,
hJtTrueBin,
jetPtBins,
xlog=True,
ylog=True,
name=filename,
proj=None,
unf2d=hReco2DProjBin,
unf2dtest=hReco2DTestProjBin,
unf=hRecoBin,
fake=hJtFakeProjBin,
start=1,
stride=1,
)
drawQA(
hJtMeas,
hJtTrue,
hJtFake,
hRecoBayes,
hRecoSVD,
hReco2DProj,
hZ,
hZTrue,
hZMeas,
hZFake,
hMultiMeas,
hMultiTrue,
hJetPt,
hJetPtTrue,
hJetPtMeas,
hRecoJetPt,
responseMatrix,
)
outFile.Close()
mc_hist = supplementaryMC.Get('PFJet_pt_m_AK8')
mc_hist.Scale(1. / mc_hist.Integral())
responses_softdrop = []
unfolded_softdrop = []
mc_hist_softdrop = supplementaryMC.Get('PFJet_pt_m_AK8SD')
mc_hist_softdrop.Scale(1. / mc_hist_softdrop.Integral())
for i in range(0, 10):
responses_softdrop.append(jackknife_mc.Get('2d_response_softdrop' +
str(i)))
responses.append(jackknife_mc.Get('2d_response' + str(i)))
unfolded_softdrop.append(
RooUnfoldBayes(responses_softdrop[i], mc_hist_softdrop, 4).Hreco())
unfolded.append(RooUnfoldBayes(responses[i], mc_hist, 4).Hreco())
unfolded_data.append(RooUnfoldBayes(responses[i], data_hist, 4).Hreco())
unfolded_data_softdrop.append(
RooUnfoldBayes(responses_softdrop[i], data_hist_softdrop, 4).Hreco())
unfolded_data[i].Scale(1. / unfolded_data[i].Integral())
unfolded_data_softdrop[i].Scale(1. / unfolded_data_softdrop[i].Integral())
unfolded_softdrop[i].Scale(1. / unfolded_softdrop[i].Integral())
unfolded[i].Scale(1. / unfolded[i].Integral())
del responses[:]
del responses_softdrop[:]
def main(optunf="Bayes"):
optunfs = ["Bayes", "SVD", "TUnfold", "Invert", "Reverse"]
if not optunf in optunfs:
txt = "Unfolding option " + optunf + " not recognised"
raise ValueError(txt)
global hReco, hMeas, hTrue, hTrueEE, hTrueMM, hMeasMM, hMeasEE
#c2 = TCanvas("c2")
print "==================================== TRAIN ===================================="
# Create response matrix object
inputdirZZ = "/afs/cern.ch/work/c/chasco/CMSSW_5_3_11/src/CMGTools/HtoZZ2l2nu/test/results/UNFOLDacc/HADD/UnfoldInput/" #ZZ
# inputdirWZ = "/afs/cern.ch/work/c/chasco/CMSSW_5_3_11/src/CMGTools/HtoZZ2l2nu/test/results/UNFOLDWZ/HADD/UnfoldInput/" #WZ
# inputdirNoPrep = "/afs/cern.ch/work/c/chasco/CMSSW_5_3_11/src/CMGTools/HtoZZ2l2nu/test/results/ZZoutput/" #data, NRB
inputdirDY = "/afs/cern.ch/work/c/chasco/CMSSW_5_3_11/src/CMGTools/HtoZZ2l2nu/DYbkgd/" #gamma
#inputGEN = "/afs/cern.ch/work/c/chasco/CMSSW_5_3_11/src/CMGTools/HtoZZ2l2nu/Unfolding/PT.root"
inputdir = "/afs/cern.ch/work/c/chasco/CMSSW_5_3_11/src/CMGTools/HtoZZ2l2nu/test/results/EWKSET/HADD/UnfoldInput/" #ZZ
inputdirWZ = inputdir #WZ
inputdirNoPrep = "/afs/cern.ch/work/c/chasco/CMSSW_5_3_11/src/CMGTools/HtoZZ2l2nu/test/results/EWKSET/" #data, NRB
#inputdirNoPrep2 = "/afs/cern.ch/work/c/chasco/CMSSW_5_3_11/src/CMGTools/HtoZZ2l2nu/test/results/" #data, NRB
inputdir2 = "/afs/cern.ch/work/c/chasco/CMSSW_5_3_11/src/CMGTools/HtoZZ2l2nu/test/results/NOSET/HADD/UnfoldInput/"
# inputdirWZ = inputdir
fw = '/afs/cern.ch/work/c/chasco/CMSSW_5_3_11/src/CMGTools/HtoZZ2l2nu/Unfolding/MCFM_withAcceptance/ZZlept_tota_MSTW200_90__90__test.root'
fwo = '/afs/cern.ch/work/c/chasco/CMSSW_5_3_11/src/CMGTools/HtoZZ2l2nu/Unfolding/MCFM_withoutAcceptance/ZZlept_tota_MSTW200_90__90__test.root'
#inputdirOTH = "/afs/cern.ch/work/c/chasco/CMSSW_5_3_11/src/CMGTools/HtoZZ2l2nu/test/results/NOSET/HADD/UnfoldInput/"
#inputdirWZOTH = inputdirOTH
#inputdirEWK = "/afs/cern.ch/work/c/chasco/CMSSW_5_3_11/src/CMGTools/HtoZZ2l2nu/test/results/EWKSET/HADD/UnfoldInput/" #ZZ
PAIR = ['zptG', 'zpt']
outputnameGR = PAIR[0] + '_vs_' + PAIR[1] + '.png'
outputnameG = PAIR[0] + '.png'
outputnameR = PAIR[1] + '.png'
cutData = "(zpt>45)*(preco>0)" #"preco"
cutGMM = "((L1GId*L2GId)==-13*13)"
cutRMM = "((l1id*l2id)==-13*13)"
cutGEE = "((L1GId*L2GId)==-11*11)"
cutREE = "((l1id*l2id)==-11*11)"
cutQQ = "(abs(l1id*l2id) == 11*13)"
hsWW = makehistos("DataFused.root", inputdirNoPrep)
hsDY = makehistosDY('gamma_out_8_MoreBins_ll.root', inputdirDY)
print "MEASURED DATA " + "=" * 30
hMeasMM = Make1DPlot(inputdirNoPrep + "DataFused.root", "finalTree",
PAIR[1], "Data.png", cutRMM + "*" + cutData)
hMeasEE = Make1DPlot(inputdirNoPrep + "DataFused.root", "finalTree",
PAIR[1], "Data.png", cutREE + "*" + cutData)
#hMeasQQ = Make1DPlot(inputdirNoPrep+"DataFused.root","finalTree",PAIR[1],"Data.png",cutQQ + "*" + cutData)
print "MEASURED DATA " + "=" * 30
print hMeasMM.Integral(), hMeasEE.Integral(), "Measured"
print "NRB and DY"
print hsWW[0].Integral(), hsWW[1].Integral(), "WW"
print hsDY[0].Integral(), hsDY[1].Integral(), "DY"
#MCFM rescaling and plotting
#
hMCFM = GetMCFMHisto(fwo, True)
hMCFM.Scale(143.2 / hMCFM.Integral())
hMCFM = REBIN(hMCFM)
print hMCFM.Integral()
hMCFM.Print("range")
#sys.exit("donesies")
MWO = GetMCFMHisto(fwo, True)
MW = GetMCFMHisto(fw, False)
MWO.Divide(MW)
MWO.Print("range")
MWO = REBIN(MWO)
MWO.Print("range")
# MWO = REBIN(GetMCFMHisto(fwo,True))
# MW = REBIN(GetMCFMHisto(fw, False))
# MWO.Divide(MW)
# MWO.Print("range")
#sys.exit("donesies")
systematic = [
'_jer', '_jes', '_umet', '_les', '_pu', '_btag', '_qcd', '_pdf'
] #'_jer','_jes','_umet','_les','_pu','_btag','_sherpa','_qcd','_pdf']
# systematic = ['_jer','_jes','_qcd','_pdf']
UD = ['up', 'down']
SYST_UD = ['']
for syst in systematic:
for ud in UD:
SYST_UD.append(syst + ud)
MCGenHistos = []
MCRecoHistos = []
hTrues = []
hDiffs = []
hCloses = []
hunfolds = []
########
MCGenHistosMM = []
MCRecoHistosMM = []
hTruesMM = []
hDiffsMM = []
hClosesMM = []
hunfoldsMM = []
########
MCGenHistosEE = []
MCRecoHistosEE = []
hTruesEE = []
hDiffsEE = []
hClosesEE = []
hunfoldsEE = []
######
MCGenHistosMM2 = []
MCGenHistosEE2 = []
MCGenHistos2 = []
for syst_ud in SYST_UD:
print "=O" * 40
print syst_ud
print "O=" * 40
SYS = "(sys" + syst_ud + ">0)"
cutR = "Eweight*XS*BR*LUM*(1/NGE)*(B2/B3)*(preco>0)*" + SYS
# cutG = "Eweight*BR*(B2/B3)*(pgen>0)*"+SYS #Accstring
# [5718.8469039797783, 1737.4891278286839, 1737.4891278286839] fb
#SSS = 2.1299863918*(3.0)
#SSS = 2.1299863918*(3.0/2.0)
#SSS = (3.0/2.0)
#SSS = (3.0)
#cutG = str(SSS)+"*Eweight*XS*BR*LUM*(1/NGE)*(B2/B3)*(pgen>0)*"+SYS #Accstring
cutG = "Eweight*XS*BR*LUM*(1/NGE)*(B2/B3)*(pgen>0)*" + SYS #Accstring
# cutG = str(SSS)+"*Eweight*XS*BR*LUM*(1/NGE)*(B2/B3)*(1/Acc2)*(pgen>0)*"+SYS #Accstring
#cutG = str(SSS)+"*Eweight*XS*BR*LUM*(1/NGE)*(B2/B3)*(pgen>0)*"+SYS #Accstring
cutGR = "Eweight*XS*BR*LUM*(1/NGE)*(B2/B3)*(pgen>0)*(preco>0)*" + SYS #+Accstring
cutWZ = "Eweight*XS*BR*LUM*(1/NGE)*(B2/B3)*(preco>0)*" + SYS
cutGEN = "XS*LUM*(1/NGE)*(B2/B3)*(g_pt>45.0)"
cutee = "(ee>0.0)"
cutmm = "(mumu>0.0)"
# MCRecoHistoMM = REBIN(Make1DPlot(inputdir+"ZZ.root","tmvatree",PAIR[1],outputnameR,cutR+"*"+cutRMM))
# MCRecoHistoEE = REBIN(Make1DPlot(inputdir+"ZZ.root","tmvatree",PAIR[1],outputnameR,cutR+"*"+cutREE))
# MCGenHistoMM = REBIN(Make1DPlot(inputdir+"ZZ.root","tmvatree",PAIR[0],outputnameG,cutG+"*"+cutGMM))
# MCGenHistoEE = REBIN(Make1DPlot(inputdir+"ZZ.root","tmvatree",PAIR[0],outputnameG,cutG+"*"+cutGEE))
MCRecoHistoMM = Make1DPlot(inputdir + "ZZ.root", "tmvatree", PAIR[1],
outputnameR, cutR + "*" + cutRMM)
MCRecoHistoEE = Make1DPlot(inputdir + "ZZ.root", "tmvatree", PAIR[1],
outputnameR, cutR + "*" + cutREE)
MCGenHistoMM = Make1DPlotGEN(inputdir + "ZZ.root", "tmvatree", PAIR[0],
outputnameG, cutG + "*" + cutGMM)
MCGenHistoEE = Make1DPlotGEN(inputdir + "ZZ.root", "tmvatree", PAIR[0],
outputnameG, cutG + "*" + cutGEE)
# MCGenHistoMM = Make1DPlotGEN(inputGEN,"GPT","g_pt",outputnameG,cutGEN+"*"+cutmm)
# MCGenHistoEE = Make1DPlotGEN(inputGEN,"GPT","g_pt",outputnameG,cutGEN+"*"+cutee)
# MCGenHistoMM.Print("range")
# sys.exit("done")
GenVsReco2DHistoMM = Make2DPlot(inputdir + "ZZ.root", "tmvatree", PAIR,
'MM' + outputnameGR,
cutGR + "*" + cutGMM + "*" + cutRMM,
syst_ud)
GenVsReco2DHistoEE = Make2DPlot(inputdir + "ZZ.root", "tmvatree", PAIR,
'EE' + outputnameGR,
cutGR + "*" + cutGEE + "*" + cutREE,
syst_ud)
# sys.exit("done")
#
MCGenHistoMM2 = Make1DPlotGEN(inputdir2 + "ZZ.root", "tmvatree",
PAIR[0], outputnameG,
cutG + "*" + cutGMM)
MCGenHistoEE2 = Make1DPlotGEN(inputdir2 + "ZZ.root", "tmvatree",
PAIR[0], outputnameG,
cutG + "*" + cutGEE)
#MCGenHistoMM2 = Make1DPlotMCFM(1)
#MCGenHistoEE2 = Make1DPlotMCFM(1)
hWZMM = Make1DPlot(inputdirWZ + "WZ.root", "tmvatree", PAIR[1],
"WZ.png", cutWZ + "*" + cutRMM) #MC
hWZEE = Make1DPlot(inputdirWZ + "WZ.root", "tmvatree", PAIR[1],
"WZ.png", cutWZ + "*" + cutREE) #MC
print hWZMM.Integral(), hWZEE.Integral(), "WZ MM EE"
print MCRecoHistoMM.Integral(), MCRecoHistoEE.Integral(), "ZZ MM EE"
BKGDSMM = [REBIN(hsWW[0]), REBIN(hsDY[0]), hWZMM]
BKGDSEE = [REBIN(hsWW[1]), REBIN(hsDY[1]), hWZEE]
for bk in (BKGDSMM + BKGDSEE):
print bk.Integral(), "subtract"
print hMeasMM.Integral()
print hMeasEE.Integral()
hDiffMM = SubtractMany1DPlots(hMeasMM, BKGDSMM)
hDiffEE = SubtractMany1DPlots(hMeasEE, BKGDSEE)
# hDiffMM = hMeasMM
# hDiffEE = hMeasEE
print hDiffMM.Integral(), quadadd(BinError(hDiffMM))
print hDiffEE.Integral(), quadadd(BinError(hDiffEE))
print "Difference between Data and Background^"
print hMeasMM.Integral(), quadadd(BinError(hMeasMM))
print hMeasEE.Integral(), quadadd(BinError(hMeasEE))
print "DATA yield^"
print REBIN(hsWW[0]).Integral(), quadadd(BinError(REBIN(hsWW[0])))
print REBIN(hsWW[1]).Integral(), quadadd(BinError(REBIN(hsWW[1])))
print "NRB yield^"
print REBIN(hsDY[0]).Integral(), quadadd(BinError(REBIN(hsDY[0])))
print REBIN(hsDY[1]).Integral(), quadadd(BinError(REBIN(hsDY[1])))
print "DY yield^"
print hWZMM.Integral(), quadadd(BinError(hWZMM))
print hWZEE.Integral(), quadadd(BinError(hWZEE))
print "WZ yield^"
print MCRecoHistoMM.Integral(), quadadd(BinError(MCRecoHistoMM))
print MCRecoHistoEE.Integral(), quadadd(BinError(MCRecoHistoEE))
print "ZZ yield^"
#sys.exit("donesies")
responseMM = RooUnfoldResponse(MCRecoHistoMM, MCGenHistoMM,
GenVsReco2DHistoMM)
responseEE = RooUnfoldResponse(MCRecoHistoEE, MCGenHistoEE,
GenVsReco2DHistoEE)
# MCRecoHistoMM.Print("range")
# MCGenHistoMM.Print("range")
# GenVsReco2DHistoMM.Print("range")
#sys.exit("DONE")
#sys.exit("done")
# print "==================================== TEST ====================================="
# #hTrue= TH1D( "true", "Test Truth", 20, -40.0, 40.0 )
# #hMeas= TH1D( "meas", "Test Measured", 40, -10.0, 10.0 )
# #hTrue= TH1D( "true", "Test Truth", 10, 40.0, 760.0 )
# # hMeas= TH1D( "meas", "Test Measured", 10, 0.0, 800.0 )
print "==================================== UNFOLD ==================================="
print "Unfolding method:", optunf
if "Bayes" in optunf:
# Bayes unfoldung with 4 iterations
#unfoldMM= RooUnfoldBayes( responseMM, hDiffMM, 4)
closureMM = RooUnfoldBayes(responseMM, MCRecoHistoMM, 4)
#unfoldEE= RooUnfoldBayes( responseEE, hDiffEE, 4)
closureEE = RooUnfoldBayes(responseEE, MCRecoHistoEE, 4)
unfoldsysMM = RooUnfoldBayes(responseMM, hDiffMM, 4)
unfoldsysEE = RooUnfoldBayes(responseEE, hDiffEE, 4)
# if syst_ud == "":
# unfoldsysMM.SetNToys(100)
# unfoldsysMM.IncludeSystematics()
# unfoldsysEE.SetNToys(100)
# unfoldsysEE.IncludeSystematics()
print "Bayes " * 20
#closure= RooUnfoldBayes( response, MCRecoHisto , 4 )
#unfold= RooUnfoldBayes( response, hMeas, 10, False, True )
elif "SVD" in optunf:
# SVD unfoding with free regularisation
# unfold= RooUnfoldSvd( response, hMeas, 20 )
unfold = RooUnfoldSvd(response, hMeas)
elif "TUnfold" in optunf:
# TUnfold with fixed regularisation tau=0.002
# unfold= RooUnfoldTUnfold( response, hMeas )
unfold = RooUnfoldTUnfold(response, hMeas, 0.002)
elif "Invert" in optunf:
unfold = RooUnfoldInvert(response, hMeas)
elif "Reverse" in optunf:
unfold = RooUnfoldBayes(response, hMeas, 1)
#hTrueMM= unfoldMM.Hreco()
hCloseMM = closureMM.Hreco()
#hTrueEE= unfoldEE.Hreco()
hCloseEE = closureEE.Hreco()
if syst_ud == "":
#unfoldsysMM.SetNToys(500)
# unfoldsysMM.IncludeSystematics(0) # 0 DATA ONLY
unfoldsysMM.IncludeSystematics(1) # 1 EVERYTHING
# unfoldsysMM.IncludeSystematics(2) # 2 MC RESPONSE
#unfoldsysEE.SetNToys(500)
unfoldsysEE.IncludeSystematics(1)
# hunfoldMM = unfoldsysMM.Hreco(2) # 2 DIRECT ERROR PROAGATION, ONLY WITH IncludeSystematics(0) -- data only. Can't directly propagate MC response unc.
hunfoldMM = unfoldsysMM.Hreco(2)
hunfoldEE = unfoldsysEE.Hreco(
2
) # 3 Toy MC error evaluation. Would apply to both data-MC uncertainty, and MC response uncertainty.
print "STUFF" * 40
hunfoldMM.Print("range")
print "UNFOLDED NUMBERS :--:" * 10
print hunfoldMM.Integral(), quadadd(BinError(hunfoldMM))
print hunfoldEE.Integral(), quadadd(BinError(hunfoldEE))
print "UNFOLDED NUMBERS :--:" * 10
#############################
#IMPORTANT!!! Includesys&Hreco: 0&2, 1&2 (some errors), N&2 (bayesian)*, 1&3, 2&3
#############################
else:
hunfoldMM = unfoldsysMM.Hreco()
hunfoldEE = unfoldsysEE.Hreco()
print " ---- TEST @@@@@@@@@@@@@@@@@@@@@@@@@ "
hunfoldEE.Print("range")
hunfoldMM.Print("range")
# print hTrueMM.Integral(), hTrueEE.Integral(), "True"
print MCGenHistoMM.Integral(), MCGenHistoEE.Integral(), "Gen"
print hunfoldMM.Integral(), hunfoldEE.Integral(), "unfold"
#makeComparison(MCGenHistoMM,MCRecoHistoMM,hTrueMM,hDiffMM,hCloseMM,"MM"+syst_ud)
#makeComparison(MCGenHistoEE,MCRecoHistoEE,hTrueEE,hDiffEE,hCloseEE,"EE"+syst_ud)
#RESCALE TO MCFM with kinematic cut ratios
hunfoldEE.Multiply(MWO)
hunfoldMM.Multiply(MWO)
MCGenHistoEE.Multiply(MWO)
MCGenHistoMM.Multiply(MWO)
MCGenHistoEE2.Multiply(MWO)
MCGenHistoMM2.Multiply(MWO)
SSS = 3.0 #3.0 for single channel, 3.0/2.0 for both
hunfoldMM.Scale(SSS)
hunfoldEE.Scale(SSS)
MCGenHistoMM.Scale(SSS)
MCGenHistoEE.Scale(SSS)
MCGenHistoMM2.Scale(SSS)
MCGenHistoEE2.Scale(SSS)
print MCGenHistoMM.Integral(), MCGenHistoEE.Integral(), "Gen"
print hunfoldMM.Integral(), hunfoldEE.Integral(), "unfold"
#sys.exit("donesies")
#For mumu channel
hDiffsEE.append([hDiffEE, syst_ud])
hunfoldsEE.append([hunfoldEE, syst_ud])
if syst_ud == "":
hClosesEE.append([hCloseEE, syst_ud])
MCGenHistosEE.append([MCGenHistoEE, syst_ud])
MCRecoHistosEE.append([MCRecoHistoEE, syst_ud])
#MCGenHistosEE2.append([hMCFM,syst_ud])
#For mumu channel
hDiffsMM.append([hDiffMM, syst_ud])
hunfoldsMM.append([hunfoldMM, syst_ud])
if syst_ud == "":
hClosesMM.append([hCloseMM, syst_ud])
MCGenHistosMM.append([MCGenHistoMM, syst_ud])
MCRecoHistosMM.append([MCRecoHistoMM, syst_ud])
MCGenHistosMM2.append([hMCFM, syst_ud])
#For combined channel
hDiffs.append([ADDER(hDiffMM, hDiffEE, 1), syst_ud])
hunfolds.append([ADDER(hunfoldMM, hunfoldEE, 1), syst_ud])
if syst_ud == "":
hCloses.append([ADDER(hCloseMM, hCloseEE, 1), syst_ud])
MCGenHistos.append([ADDER(MCGenHistoMM, MCGenHistoEE, 1), syst_ud])
MCRecoHistos.append(
[ADDER(MCRecoHistoMM, MCRecoHistoEE, 1), syst_ud])
MCGenHistos2.append(
[ADDER(MCGenHistoMM2, MCGenHistoEE2, 1), syst_ud])
print "=|" * 20
print syst_ud * 20
print "unfold yield:"
print(hunfoldMM.Integral() + hunfoldEE.Integral()) / 19.7
print(MCGenHistoMM.Integral() + MCGenHistoEE.Integral()) / 19.7, "Gen"
hunfoldMM.Print("range")
hunfoldEE.Print("range")
print ">GEN<" * 40
MCGenHistoMM.Print("range")
MCGenHistoEE.Print("range")
print "=|" * 20
# print "=|"*20
# print "unfold yield:"
# print (hunfoldMM.Integral()+hunfoldEE.Integral())/19.7
# print "=|"*20
# hDiffs.append([hDiffEE,syst_ud])
# hunfolds.append([hunfoldEE,syst_ud])
# if syst_ud == "":
# hCloses.append([hCloseEE,syst_ud])
# MCGenHistos.append([MCGenHistoEE,syst_ud])
# MCRecoHistos.append([MCRecoHistoEE,syst_ud])
#QQQ = makeComparison(LUMscale(MCGenHistos),LUMscale(MCRecoHistos),LUMscale(hunfolds),LUMscale(hDiffs),LUMscale(MCGenHistos2),"comb")
#QQQ= makeComparison(LUMscale(MCGenHistosEE),LUMscale(MCRecoHistosEE),LUMscale(hunfoldsEE),LUMscale(hDiffsEE),MCGenHistosMM2,"ee")
#QQQ = makeComparison(LUMscale(MCGenHistosMM),LUMscale(MCRecoHistosMM),LUMscale(hunfoldsMM),LUMscale(hDiffsMM),MCGenHistosMM2,"mm")
#QQQ = makeComparison(LUMscale(MCGenHistos),LUMscale(MCRecoHistos),LUMscale(hunfolds),LUMscale(hDiffs),MCGenHistosMM2,"comb")
QQQ = makeComparison(LUMscale(MCGenHistosEE), LUMscale(MCRecoHistosEE),
LUMscale(hunfoldsEE), LUMscale(hDiffsEE),
MCGenHistosMM2, "ee")
#QQQ = makeComparison(LUMscale(MCGenHistosMM),LUMscale(MCRecoHistosMM),LUMscale(hunfoldsMM),LUMscale(hDiffsMM),MCGenHistosMM2,"mm")
Neemm = QQQ[0]
#Nee = makeComparison(LUMscale(MCGenHistosEE),LUMscale(MCRecoHistosEE),LUMscale(hunfoldsEE),LUMscale(hDiffsEE),LUMscale(hClosesEE),"ee")
#Nmm = makeComparison(LUMscale(MCGenHistosMM),LUMscale(MCRecoHistosMM),LUMscale(hunfoldsMM),LUMscale(hDiffsMM),LUMscale(hClosesMM),"mm")
# #makeComparison(MCGenHistos,MCRecoHistos,hunfolds,hDiffs,hCloses)
# print len(MCRecoHistos)
# print len(hunfolds)
# print "Bin Content Check"
# print "#"*30
# print "#"*10, "DIMUON CHANNEL", "#"*10
# print "MCReco", MCRecoHistosMM[0][0].Integral(), BinContent(MCRecoHistosMM[0][0])
# print "MCGen", MCGenHistosMM[0][0].Integral(), BinContent(MCGenHistosMM[0][0])
# print "True", hunfoldsMM[0][0].Integral(), BinContent(hunfoldsMM[0][0])
# print "Diff", hDiffsMM[0][0].Integral(), BinContent(hDiffsMM[0][0])
# print "#"*30
# print "#"*10, "DIELECTRON CHANNEL", "#"*10
# print "MCReco", MCRecoHistosEE[0][0].Integral(), BinContent(MCRecoHistosEE[0][0])
# print "MCGen", MCGenHistosEE[0][0].Integral(), BinContent(MCGenHistosEE[0][0])
# print "True", hunfoldsEE[0][0].Integral(), BinContent(hunfoldsEE[0][0])
# print "Diff", hDiffsEE[0][0].Integral(), BinContent(hDiffsEE[0][0])
# print "#"*30
# print "#"*10, "COMBINED CHANNEL", "#"*10
# print "MCReco", MCRecoHistos[0][0].Integral(), BinContent(MCRecoHistos[0][0])
# print "MCGen", MCGenHistos[0][0].Integral(), BinContent(MCGenHistos[0][0])
# print "True", hunfolds[0][0].Integral(), BinContent(hunfolds[0][0])
# print "Diff", hDiffs[0][0].Integral(), BinContent(hDiffs[0][0])
# print "#"*30
# print "#"*30
# print "#"*30
# print hunfoldsMM[0][0].Integral()*(3.0)*2.13, "ZZ->2l2nu cross section, muon"
# print hunfoldsEE[0][0].Integral()*(3.0)*2.13, "ZZ->2l2nu cross section, electron"
# print hunfolds[0][0].Integral()*(3.0/2.0)*2.13, "ZZ->2l2nu cross section, combined"
# print "#"*30
# print Nmm
# print Nee
print Neemm
print QQQ[1]
#print Nmm
#print Nee
# print "@"*30
# print "@"*30
# print "@"*30
# S=2.1299863918
# dS=0.0451194907919
# print numpy.multiply(Nmm,S*3)
# print numpy.multiply(Nee,S*3)
# print numpy.multiply(Neemm,S*3.0/2.0)
# print "#"*30
# print "#"*30
# print "#"*30
# print numpy.multiply(RescaleToPreZpt45(Nmm,S,dS),3.0), "muons"
# print numpy.multiply(RescaleToPreZpt45(Nee,S,dS),3.0), "electrons"
# print numpy.multiply(RescaleToPreZpt45(Neemm,S,dS),3.0/2.0), "combined"
# Just unfolding
# [ 317.58667967 124.62000891 124.62000891] muons
# [ 204.44743361 93.817413 93.817413 ] electrons
# [ 261.01705568 78.68129176 78.68129176] combined
#all else
# Bin Content Check
# ##############################
# ########## DIMUON CHANNEL ##########
# MCReco 5.48821856594 [1.6566265821456909, 1.5307177305221558, 2.0848486423492432, 0.21176119148731232, 0.0042644194327294827]
# MCGen 50.3628177345 [28.911533355712891, 8.1944065093994141, 11.555961608886719, 1.5890809297561646, 0.11183533072471619]
# True 49.7008933779 [27.718360900878906, 7.7494301795959473, 13.04161262512207, 1.2007522583007812, -0.0092625860124826431]
# Diff 5.54152165353 [1.6073417663574219, 1.4832497835159302, 2.371938943862915, 0.16530285775661469, -0.086311697959899902]
# ##############################
# ########## DIELECTRON CHANNEL ##########
# MCReco 3.674643751 [1.0442177057266235, 0.9972614049911499, 1.4791457653045654, 0.15098364651203156, 0.0030352284666150808]
# MCGen 33.3869778588 [18.415399551391602, 5.4404187202453613, 8.2116708755493164, 1.2221240997314453, 0.097364611923694611]
# True 31.9951079488 [22.219881057739258, 3.5517585277557373, 5.7815923690795898, 0.44187599420547485, 0.0]
# Diff 3.02480854467 [1.2913563251495361, 0.70207256078720093, 1.0841209888458252, 0.059399198740720749, -0.11214052885770798]
# ##############################
# ########## COMBINED CHANNEL ##########
# MCReco 9.16539874254 [2.7004456520080566, 2.5281918048858643, 3.5664489269256592, 0.36300751566886902, 0.0073048430494964123]
# MCGen 83.7497940361 [47.326930999755859, 13.634824752807617, 19.767633438110352, 2.8112049102783203, 0.2091999351978302]
# True 81.6960010286 [49.938240051269531, 11.301189422607422, 18.823205947875977, 1.6426281929016113, -0.0092625860124826431]
# Diff 8.56633037329 [2.8986983299255371, 2.1853222846984863, 3.4560599327087402, 0.22470206022262573, -0.19845223426818848]
# ##############################
# ##############################
# ##############################
# 317.588708685 ZZ->2l2nu cross section, muon
# 204.448739793 ZZ->2l2nu cross section, electron
# 261.018723287 ZZ->2l2nu cross section, combined
# ##############################
# [49.700893377885222, 24.356035601246898, 24.954409212519852]
# [31.99510794878006, 17.444453802993618, 19.251802135458732]
# [81.696001028642058, 34.234254955780528, 37.056285190785381]
# ##############################
# ##############################
# ##############################
# [ 317.58667967 155.77940538 159.59950658] muons 93.47233023637472982875, 99.70885557387058841952 quaddifference to just unfolding
# [ 204.44743361 111.55344624 123.09443832] electrons 60.35283246052335146028, 79.69023631100434516015
# [ 261.01705568 109.51740717 118.52311228] combined 76.17950380658385143785, 88.63962134122213690629
# os.system(str(Npm)+' > output.txt')
# SystematicErrors(MCRecoHistos,MCGenHistos,hTrues,hDiffs,hCloses)
#
#testtest
return
xt = gRandom.BreitWigner(0.3, 2.5)
x = smear(xt)
if x != None:
response.Fill(x, xt)
else:
response.Miss(xt)
print "==================================== TEST ====================================="
hTrue = TH1D("true", "Test Truth", 40, -10.0, 10.0)
hMeas = TH1D("meas", "Test Measured", 40, -10.0, 10.0)
# Test with a Gaussian, mean 0 and width 2.
for i in xrange(10000):
xt = gRandom.Gaus(0.0, 2.0)
x = smear(xt)
hTrue.Fill(xt)
if x != None: hMeas.Fill(x)
print "==================================== UNFOLD ==================================="
unfold = RooUnfoldBayes(response, hMeas, 4)
# OR
# unfold= RooUnfoldSvd (response, hMeas, 20); # OR
# unfold= RooUnfoldTUnfold (response, hMeas); # OR
# unfold= RooUnfoldIds (response, hMeas, 3); # OR
hReco = unfold.Hreco()
unfold.PrintTable(cout, hTrue)
hReco.Draw()
hMeas.Draw("SAME")
hTrue.SetLineColor(8)
hTrue.Draw("SAME")
else:
chiHists1d.append(
Proj2D_Y(Reco2d, minMass, maxMass, Reco2d.GetName(), True))
if compToGen:
chiTests1d.append(
Proj2D_Y(Gen2d, minMass, maxMass, Gen2d.GetName(), True))
else:
chiTests1d.append(
Proj2D_Y(Reco2d, minMass, maxMass, Reco2d.GetName(), True))
## now do unfolding
print "CCLA: Unfolding ", Reco2d, "using matix ", response
if DAgostini == True:
unfold2d = RooUnfoldBayes(response, Reco2d, NITER)
else:
unfold2d = RooUnfoldInvert(response, Reco2d)
unfold2d.Print()
new_hists.append(unfold2d)
unfold2d.SetNToys(30000)
hUnf2d = unfold2d.Hreco(2)
hErr2d = unfold2d.Ereco(2)
#chi2 = RooUnfold.Chi2 (unfold2d,2)
chi2comp = Proj2D_Y(Gen2d, minMass, maxMass, Gen2d.GetName(), True)
def unfold(self):
if self._fillFakes:
fakes = self._hJtFake2D
else:
fakes = None
if self._hJtTrue2D:
self._response2D = make2Dresponse(
self._2Dresponses,
self._jetPtBins,
self._hJtMeas2D,
self._hJtTrue2D,
misses=self._misses2D,
fakes=fakes,
)
else:
self._response2D = make2Dresponse(
self._2Dresponses,
self._jetPtBins,
self._hJtMeas2D,
self._hJtMeas2D,
misses=self._misses2D,
fakes=fakes,
)
if not self._fillFakes:
print("Scale hJtFake2D by {}".format(1.0 *
sum(self._numberJetsMeasBin) /
self._numberJetsMeasTrain))
self._hJtFake2D.Scale(1.0 * sum(self._numberJetsMeasBin) /
self._numberJetsMeasTrain)
self._hJtMeas2D.Add(self._hJtFake2D, -1)
self._hJtFakeProjBin = [
makeHist(
self._hJtFake2D.ProjectionX("histFake{}".format(i), i, i),
bins=self._LogBinsJt,
) for i in range(1, len(self._jetBinBorders))
]
for h, nj in zip(self._hJtFakeProjBin, self._numberJetsMeasBin):
if nj > 0:
h.Scale(1.0 / nj, "width")
if self._responseJetPtCoarse is None:
print(self._responseJetPtCoarse)
self._responseJetPtCoarse = createResponse(
self._hJetPtMeasCoarse, self._hresponseJetPtCoarse)
if self._responseJetPt is None:
self._responseJetPt = createResponse(self._hJetPtMeas,
self._hresponseJetPt)
self._hJetPtRecoCoarse = unfoldJetPt(self._hJetPtMeasCoarse,
self._responseJetPtCoarse,
self._jetBinBorders)
self._hJetPtReco = unfoldJetPt(self._hJetPtMeas, self._responseJetPt,
self._LogBinsX)
self._numberJetsMeasFromHist = [
self._hJetPtMeasCoarse.GetBinContent(i)
for i in range(1,
self._hJetPtMeasCoarse.GetNbinsX() + 1)
]
if not self._IsData:
self._numberJetsTrueFromHist = [
self._hJetPtTrueCoarse.GetBinContent(i)
for i in range(1,
self._hJetPtTrueCoarse.GetNbinsX() + 1)
]
self._numberJetsFromReco = [
self._hJetPtRecoCoarse.GetBinContent(i)
for i in range(1, self._hJetPtRecoCoarse.GetNbinsX())
]
self.printJetNumbers()
unfold2D = RooUnfoldBayes(self._response2D, self._hJtMeas2D,
self._Niter)
self._hJtReco2D = make2DHist(unfold2D.Hreco(),
xbins=self._LogBinsJt,
ybins=self._jetBinBorders)
self._hJtMeasBin = [
makeHist(
self._hJtMeas2D.ProjectionX("histMeas{}".format(i), i, i),
bins=self._LogBinsJt,
) for i in range(1, len(self._jetBinBorders))
]
if not self._IsData:
self._hJtTestMeasBin = [
makeHist(
self._hJtTestMeas2D.ProjectionX("histTestMeas{}".format(i),
i, i),
bins=self._LogBinsJt,
) for i in range(1, len(self._jetBinBorders))
]
self._hJtTrueBin = [
makeHist(
self._hJtTrue2D.ProjectionX("histMeas{}".format(i), i, i),
bins=self._LogBinsJt,
) for i in range(1, len(self._jetBinBorders))
]
self._hJtTestTrueBin = [
makeHist(
self._hJtTestTrue2D.ProjectionX("histMeas{}".format(i), i,
i),
bins=self._LogBinsJt,
) for i in range(1, len(self._jetBinBorders))
]
self._hJtRecoBin = [
makeHist(
self._hJtReco2D.ProjectionX("histMeas{}".format(i), i, i),
bins=self._LogBinsJt,
) for i in range(1, len(self._jetBinBorders))
]
for h, n, in zip(
self._hJtMeasBin,
self._numberJetsMeasFromHist,
):
if n > 0:
h.Scale(1.0 / n, "width")
if not self._IsData:
for h, h2, h3, n, n2, n3 in zip(
self._hJtTrueBin,
self._hJtTestTrueBin,
self._hJtTestMeasBin,
self._numberJetsTrueFromHist,
self._numberJetsTestTrueBin,
self._numberJetsTestMeasBin,
):
if n > 0:
h.Scale(1.0 / n, "width")
if n2 > 0:
h2.Scale(1.0 / n2, "width")
if n3 > 0:
h3.Scale(1.0 / n3, "width")
for h, n in zip(self._hJtRecoBin, self._numberJetsFromReco):
if n > 0:
h.Scale(1.0 / n, "width")
#####################
pvalues = array('f', [])
n_iterations = array('f', [])
y_error = array('f', [])
x_error = array('f', [])
c_refoldeds = []
refolded_hists = []
c_sanitys = []
sanity_hists = []
for i in range(1, 20, 1):
print '--------- i = ', i
unfoldreco = RooUnfoldBayes(response, reco, i)
reco_unfolded = unfoldreco.Vreco()
reco_asvector = unfoldreco.Vmeasured()
m = response.Mresponse()
print 'response : ', m.GetNrows(), ' x ', m.GetNcols()
refolded = ROOT.TVectorD(reco_unfolded)
refolded *= m #TMatrixD(m,TMatrixD.kMult,reco_unfolded)
print 'refolded : ', refolded.GetNrows()
refolded_hist = ROOT.TH1D("refolded_hist_" + str(i),
"refolded_hist_" + str(i),
reco.GetNbinsX() * reco.GetNbinsY(), 0,
reco.GetNbinsX() * reco.GetNbinsY())
sanity_hist = ROOT.TH1D("sanity_hist_" + str(i), "sanity_hist_" + str(i),
reco.GetNbinsX() * reco.GetNbinsY(), 0,
reco.GetNbinsX() * reco.GetNbinsY())
for ibin in xrange(0, refolded.GetNrows()):
def MyRooUnfold(matrix_name=args.h_matrix, h_reco_getG0_name=args.h_data, h_ptcl_getG0_name = args.h_particle,h_reco_get_bkg_name = args.h_background,outputname=args.h_data+"_unfolded",nrebin = args.nrebin):
rfile_data = TFile(args.rfile_data, 'read')
rfile_particle = TFile(args.rfile_particle, 'read')
rfile_matrix = TFile(args.rfile_matrix, 'read')
rfile_background = TFile(args.rfile_background, 'read')
myfbu = fbu.PyFBU()
myfbu.verbose = True
#GET DATA
h_reco_get = rfile_data.Get(h_reco_getG0_name)
h_reco_get.Rebin(nrebin)
#GET PARTICLE
h_ptcl_get = rfile_particle.Get(h_ptcl_getG0_name)
h_ptcl_get.Rebin(nrebin)
#GET MATRIX
h_response_unf = rfile_matrix.Get(matrix_name)
h_response_unf.ClearUnderflowAndOverflow()
h_response_unf.GetXaxis().SetRange(1, h_response_unf.GetXaxis().GetNbins() )
h_response_unf.GetYaxis().SetRange(1, h_response_unf.GetYaxis().GetNbins() )
h_response_unf.Rebin2D(nrebin,nrebin)
h_response_unf.SetName("Migration_Matrix_simulation")
########### ACCEPTANCY
h_acc = h_response_unf.ProjectionX("reco_recoandparticleX") # Reco M
h_acc.Divide(h_reco_get)
########### AKCEPTANCE saved in h_acc #############
########### EFFICIENCY
h_eff = h_response_unf.ProjectionY("reco_recoandparticleY") # Ptcl M
h_eff.Divide(h_ptcl_get)
h_reco_get_input = rfile_data.Get(h_reco_getG0_name)
h_reco_get_bkg = rfile_background.Get(h_reco_get_bkg_name)
h_reco_get_bkg.Rebin(nrebin)
h_reco_get_input_clone=h_reco_get_input.Clone("")
h_reco_get_input_clone.Add(h_reco_get_bkg,-1)
h_reco_get_input_clone.Multiply(h_acc)
h_reco_or = rfile_data.Get(h_reco_getG0_name)
h_ptcl_or = rfile_particle.Get(h_ptcl_getG0_name)
h_ptcl_or.SetMaximum(h_ptcl_or.GetMaximum()*1.5)
### ROOUNFOLD METHOD ###
m_RooUnfold = RooUnfoldBayes()
m_RooUnfold.SetRegParm( 4 )
m_RooUnfold.SetNToys( 10000 )
m_RooUnfold.SetVerbose( 0 )
m_RooUnfold.SetSmoothing( 0 )
response = RooUnfoldResponse(None, None, h_response_unf, "response", "methods")
m_RooUnfold.SetResponse( response )
m_RooUnfold.SetMeasured( h_reco_get_input_clone )
### SVD METHOD ###
m_RooUnfold_svd = RooUnfoldSvd (response, h_reco_get_input_clone, int(round(h_reco_get_input_clone.GetNbinsX()/2.0,0))) #8
svd_par = int(round(h_reco_get_input_clone.GetNbinsX()/2.0,0))
m_RooUnfold_T = RooUnfoldTUnfold (response, h_reco_get_input_clone) # OR
m_RooUnfold_Ids= RooUnfoldIds (response, h_reco_get_input_clone,int(round(h_reco_get_input_clone.GetNbinsX()/12.0,0))) ## TO DO, SET PARAMETERS TO THE BINNING
Ids_par = int(round(h_reco_get_input_clone.GetNbinsX()/12.0,0))
### FBU METHOD ###
h_response_unf_fbu = TransposeMatrix(h_response_unf)
h_response_unf_fbu_norm = NormalizeResponse(h_response_unf_fbu)
h_response_unf_fbu_norm.SetName("Migration_Matrix_simulation_transpose")
histograms.append(h_response_unf_fbu_norm)
myfbu.response = MakeListResponse(h_response_unf_fbu_norm)
myfbu.data = MakeListFromHisto(h_reco_get_input_clone)
myfbu.lower = []
myfbu.upper = []
h_det_div_ptcl=h_reco_get_input_clone.Clone("")
h_det_div_ptcl.Divide(h_ptcl_or)
h_det_div_ptcl.Divide(h_eff)
h_det_div_ptcl.SetName("det_div_ptcl")
histograms.append(h_det_div_ptcl)
for l in range(len(myfbu.data)):
if ( args.SplitFromBinLow != 0) and ( l+1 <= args.SplitFromBinLow ):
myfbu.lower.append(h_reco_get_input_clone.GetBinContent(l+1)*(2-args.ParameterSplitFromBinLow)*h_det_div_ptcl.GetBinContent(l+1))
myfbu.upper.append(h_reco_get_input_clone.GetBinContent(l+1)*args.ParameterSplitFromBinLow*h_det_div_ptcl.GetBinContent(l+1))
elif ( args.SplitFromBinHigh != 0 ) and ( l+1 >= args.SplitFromBinHigh ):
myfbu.lower.append(h_reco_get_input_clone.GetBinContent(l+1)*(2-args.ParameterSplitFromBinHigh)*h_det_div_ptcl.GetBinContent(l+1))
myfbu.upper.append(h_reco_get_input_clone.GetBinContent(l+1)*args.ParameterSplitFromBinHigh*h_det_div_ptcl.GetBinContent(l+1))
else:
myfbu.lower.append(h_reco_get_input_clone.GetBinContent(l+1)*(2-args.par)*h_det_div_ptcl.GetBinContent(l+1))
myfbu.upper.append(h_reco_get_input_clone.GetBinContent(l+1)*args.par*h_det_div_ptcl.GetBinContent(l+1))
#myfbu.regularization = Regularization('Tikhonov',parameters=[{'refcurv':0.1,'alpha':0.2}]) works for old FBU package and python2.7 and old pymc
myfbu.run()
trace = myfbu.trace
traceName = 'Posterior_1_iteration'
posteriors_diag = MakeTH1Ds(trace, traceName)
h_reco_unfolded, h_reco_unfolded_Mean = MakeUnfoldedHisto(h_reco_or, posteriors_diag)
PlotPosteriors(posteriors_diag,outputname+'_iteration_1')
# EFFICIENCY AND ACCEPTANCY CORRECTIONS
h_reco_unfolded.Divide(h_eff)
h_reco_unfolded_Mean.Divide(h_eff)
h_reco_unfolded_roof = m_RooUnfold.Hreco()
h_reco_unfolded_roof.Divide(h_eff)
h_reco_unfolded_svd = m_RooUnfold_svd.Hreco()
h_reco_unfolded_svd.Divide(h_eff)
h_reco_unfolded_T = m_RooUnfold_T.Hreco()
h_reco_unfolded_T.Divide(h_eff)
h_reco_unfolded_Ids = m_RooUnfold_Ids.Hreco()
h_reco_unfolded_Ids.Divide(h_eff)
PlotRatio(h_reco_unfolded_Mean, h_ptcl_or, h_reco_unfolded_roof, h_reco_unfolded_svd, h_reco_unfolded_T,h_reco_unfolded_Ids, svd_par, Ids_par, outputname+'_iteration_1')
Repeat = True
j = 2
while Repeat:
print("Runnig iteration number: ",j)
myfbu.lower = []
myfbu.upper = []
for l in range(len(myfbu.data)):
posteriors_diag[l].Fit("gaus")
fit = posteriors_diag[l].GetFunction("gaus")
p1 = fit.GetParameter(1)
p2 = fit.GetParameter(2)
myfbu.lower.append(p1-4*p2)
myfbu.upper.append(p1+4*p2)
myfbu.run()
trace = myfbu.trace
traceName = 'Posterior_'+str(j)+'_iteration'
posteriors_diag = MakeTH1Ds(trace, traceName)
h_reco_unfolded, h_reco_unfolded_Mean = MakeUnfoldedHisto(h_reco_or, posteriors_diag)
Repeat = PlotPosteriors(posteriors_diag,outputname+'_iteration_'+str(j))
# EFFICIENCY AND ACCEPTANCY CORRECTIONS
h_reco_unfolded.Divide(h_eff)
h_reco_unfolded_Mean.Divide(h_eff)
h_reco_unfolded_roof = m_RooUnfold.Hreco()
h_reco_unfolded_roof.Divide(h_eff)
h_reco_unfolded_svd = m_RooUnfold_svd.Hreco()
h_reco_unfolded_svd.Divide(h_eff)
h_reco_unfolded_T = m_RooUnfold_T.Hreco()
h_reco_unfolded_T.Divide(h_eff)
h_reco_unfolded_Ids = m_RooUnfold_Ids.Hreco()
h_reco_unfolded_Ids.Divide(h_eff)
PlotRatio(h_reco_unfolded_Mean, h_ptcl_or, h_reco_unfolded_roof, h_reco_unfolded_svd, h_reco_unfolded_T,h_reco_unfolded_Ids, svd_par, Ids_par, outputname+'_iteration_'+str(j))
if j == args.maxiterations:
break
j = j+1
h_reco_unfolded.SetName("result_fbu_fit")
histograms.append(h_reco_unfolded)
h_reco_unfolded_Mean.SetName("result_fbu_Mean")
histograms.append(h_reco_unfolded_Mean)
h_reco_unfolded_roof.SetName("result_roof")
histograms.append(h_reco_unfolded_roof)
h_reco_unfolded_svd.SetName("result_svd")
histograms.append(h_reco_unfolded_svd)
h_reco_unfolded_T.SetName("result_T")
histograms.append(h_reco_unfolded_T)
h_reco_unfolded_Ids.SetName("result_Ids")
histograms.append(h_reco_unfolded_Ids)
h_eff.SetName("efficiency")
histograms.append(h_eff)
h_acc.SetName("acceptancy")
histograms.append(h_acc)
h_reco_or.SetName("reco")
histograms.append(h_reco_or)
h_ptcl_or.SetName("ptcl_simulation")
histograms.append(h_ptcl_or)
h_ratio = h_reco_unfolded.Clone("")
h_ratio.Divide(h_ptcl_or)
h_ratio.SetName("ratio_fbu_fit")
histograms.append(h_ratio)
h_ratio = h_reco_unfolded_Mean.Clone("")
h_ratio.Divide(h_ptcl_or)
h_ratio.SetName("ratio_fbu_Mean")
histograms.append(h_ratio)
h_ratio = h_reco_unfolded_roof.Clone("")
h_ratio.Divide(h_ptcl_or)
h_ratio.SetName("ratio_roof")
histograms.append(h_ratio)
h_ratio = h_reco_unfolded_svd.Clone("")
h_ratio.Divide(h_ptcl_or)
h_ratio.SetName("ratio_svd")
histograms.append(h_ratio)
m_RooUnfold_svd.PrintTable (cout, h_ptcl_or)
m_RooUnfold.PrintTable (cout, h_ptcl_or)
# CORRECTIONS TO GET CROSS SECTION
#DivideBinWidth(h_reco_unfolded_Mean)
#DivideBinWidth(h_reco_unfolded_roof)
#DivideBinWidth(h_reco_unfolded_svd)
#DivideBinWidth(h_reco_unfolded_T)
#DivideBinWidth(h_reco_unfolded_Ids)
#DivideBinWidth(h_ptcl_or)
#Lumi = 36.1e3
#for j in range(1,h_reco_unfolded_Mean.GetXaxis().GetNbins()+1):
# h_reco_unfolded_Mean.SetBinContent(j,h_reco_unfolded_Mean.GetBinContent(j)/(Lumi))
# h_reco_unfolded_Mean.SetBinError(j,h_reco_unfolded_Mean.GetBinError(j)/(Lumi))
# h_reco_unfolded_roof.SetBinContent(j,h_reco_unfolded_roof.GetBinContent(j)/(Lumi))
# h_reco_unfolded_roof.SetBinError(j,h_reco_unfolded_roof.GetBinError(j)/(Lumi))
# h_reco_unfolded_svd.SetBinContent(j,h_reco_unfolded_svd.GetBinContent(j)/(Lumi))
# h_reco_unfolded_svd.SetBinError(j,h_reco_unfolded_svd.GetBinError(j)/(Lumi))
# h_reco_unfolded_T.SetBinContent(j,h_reco_unfolded_T.GetBinContent(j)/(Lumi))
# h_reco_unfolded_T.SetBinError(j,h_reco_unfolded_T.GetBinError(j)/(Lumi))
# h_reco_unfolded_Ids.SetBinContent(j,h_reco_unfolded_Ids.GetBinContent(j)/(Lumi))
# h_reco_unfolded_Ids.SetBinError(j,h_reco_unfolded_Ids.GetBinError(j)/(Lumi))
# h_ptcl_or.SetBinContent(j,h_ptcl_or.GetBinContent(j)/(Lumi))
# h_ptcl_or.SetBinError(j,h_ptcl_or.GetBinError(j)/(Lumi))
#h_reco_unfolded_Mean_clone=h_reco_unfolded_Mean.Clone("FBU_cross_section")
#h_reco_unfolded_roof_clone=h_reco_unfolded_roof.Clone("DAgostini_cross_section")
#h_reco_unfolded_svd_clone=h_reco_unfolded_svd.Clone("SVD_cross_section")
#h_reco_unfolded_T_clone=h_reco_unfolded_T.Clone("TUnfold_cross_section")
#h_reco_unfolded_Ids_clone=h_reco_unfolded_Ids.Clone("Ids_cross_section")
#h_ptcl_or_clone=h_ptcl_or.Clone("Truth_cross_section")
#
#print("CONTROL*******************************************************************: ",h_reco_unfolded_Mean_clone.GetXaxis().GetNbins(),h_reco_unfolded_roof_clone.GetXaxis().GetNbins(),h_reco_unfolded_svd_clone.GetXaxis().GetNbins(),h_reco_unfolded_T_clone.GetXaxis().GetNbins(),h_reco_unfolded_Ids_clone.GetXaxis().GetNbins(),h_ptcl_or_clone.GetXaxis().GetNbins())
#histograms.append(h_reco_unfolded_Mean_clone)
#histograms.append(h_reco_unfolded_roof_clone)
#histograms.append(h_reco_unfolded_svd_clone)
#histograms.append(h_reco_unfolded_T_clone)
#histograms.append(h_reco_unfolded_Ids_clone)
#histograms.append(h_ptcl_or_clone)
SaveHistograms(outputname)
def unfold_mc(self):
logging.debug('Unfolding monte carlo sample')
regularisation = int(self.params['regularisation'].m)
unfold_bg = self.params['unfold_bg'].value
unfold_eff = self.params['unfold_eff'].value
unfold_unweighted = self.params['unfold_unweighted'].value
# Split data into signal, bg and all (signal+bg)
signal_data, bg_data, all_data = self.split_data()
# Load generator level data for signal
gen_data = self.load_gen_data()
# Return true map is regularisation is set to 0
if regularisation == 0:
logging.info('Returning baseline MapSet')
true = roounfold._histogram(events=gen_data,
binning=self.true_binning,
weights=gen_data['pisa_weight'],
errors=True,
name=self.output_str)
return MapSet([true])
# Get the inversed efficiency histogram
if not unfold_eff:
inv_eff = self.get_inv_eff(signal_data, gen_data)
# Set the reco and true data based on cfg file settings
reco_norm_data = None
true_norm_data = None
data = signal_data
if unfold_bg:
reco_norm_data = all_data
if unfold_eff:
true_norm_data = gen_data
if reco_norm_data is None:
reco_norm_data = signal_data
if true_norm_data is None:
true_norm_data = signal_data
if unfold_unweighted:
ones = np.ones(reco_norm_data['pisa_weight'].shape)
reco_norm_data['pisa_weight'] = ones
true_norm_data['pisa_weight'] = ones
data['pisa_weight'] = ones
# Create response object
response = self.create_response(reco_norm_data, true_norm_data, data)
# Make pseduodata
all_hist = self._histogram(events=all_data,
binning=self.reco_binning,
weights=all_data['pisa_weight'],
errors=False,
name='all',
tex=r'\rm{all}')
seed = int(self.params['stat_fluctuations'].m)
if seed != 0:
if self.random_state is None or seed != self.seed:
self.seed = seed
self.random_state = get_random_state(seed)
all_hist = all_hist.fluctuate('poisson', self.random_state)
else:
self.seed = None
self.random_state = None
all_hist.set_poisson_errors()
# Background Subtraction
if unfold_bg:
reco = all_hist
else:
bg_hist = self.get_bg_hist(bg_data)
reco = all_hist - bg_hist
reco.name = 'reco_signal'
reco.tex = r'\rm{reco_signal}'
r_flat = roounfold._flatten_to_1d(reco)
r_th1d = convert_to_th1d(r_flat, errors=True)
# Find optimum value for regularisation parameter
if self.params['optimize_reg'].value:
chisq = None
for r_idx in range(regularisation):
unfold = RooUnfoldBayes(response, r_th1d, r_idx + 1)
unfold.SetVerbose(0)
idx_chisq = unfold.Chi2(self.t_th1d, 1)
if chisq is None:
pass
elif idx_chisq > chisq:
regularisation = r_idx
break
chisq = idx_chisq
# Unfold
unfold = RooUnfoldBayes(response, r_th1d, regularisation)
unfold.SetVerbose(0)
unfolded_flat = unfold.Hreco(1)
unfold_map = unflatten_thist(in_th1d=unfolded_flat,
binning=self.true_binning,
name=self.output_str,
errors=True)
# Efficiency correction
if not unfold_eff:
unfold_map *= inv_eff
del r_th1d
del unfold
logging.info('Unfolded reco sum {0}'.format(
np.sum(unp.nominal_values(unfold_map.hist))))
return unfold_map
def unfoldJetPt(meas, response, xbins):
unfold = RooUnfoldBayes(response, meas)
unfold.SetVerbose(0)
hReco = makeHist(unfold.Hreco(), bins=xbins)
return hReco
def main():
gROOT.SetBatch()
#range for |t|
ptmin = 0.
ptmax = 0.109 # 0.109 0.01 for interference range
#default binning
ptbin = 0.004 # 0.004 0.0005 for interference range
#long bins at high |t|
ptmid = 0.06 # 0.08, value > ptmax will switch it off 0.06
ptlon = 0.01 # 0.01
#short bins at low |t|
ptlow = 0.01
ptshort = 0.0005
#mass interval
mmin = 2.8
mmax = 3.2
#dy = 2. # rapidity interval, for integrated sigma
dy = 1.
ngg = 131 # number of gamma-gamma from mass fit
lumi = 13871.907 # lumi in inv. ub
#correction to luminosity for ana/triggered events
ratio_ana = 3420950. / 3694000
#scale the lumi for |z| around nominal bunch crossing
ratio_zdc_vtx = 0.502
Reta = 0.503 # pseudorapidity preselection
#Reta = 1.
trg_eff = 0.67 # bemc trigger efficiency
ratio_tof = 1.433 # tof correction to efficiency
bbceff = 0.97 # BBC veto inefficiency
zdc_acc = 0.49 # ZDC acceptance to XnXn 0.7
#zdc_acc = 1.
br = 0.05971 # dielectrons branching ratio
#data
basedir = "../../../star-upc-data/ana/muDst/muDst_run1/sel5"
infile = "ana_muDst_run1_all_sel5z.root"
#MC
basedir_sl = "../../../star-upc-data/ana/starsim/slight14e/sel5"
#infile_sl = "ana_slight14e1x2_s6_sel5z.root"
infile_sl = "ana_slight14e1x3_s6_sel5z.root"
#
basedir_sart = "../../../star-upc-data/ana/starsim/sartre14a/sel5"
infile_sart = "ana_sartre14a1_sel5z_s6_v2.root"
#
basedir_bgen = "../../../star-upc-data/ana/starsim/bgen14a/sel5"
infile_bgen = "ana_bgen14a1_v0_sel5z_s6.root"
#infile_bgen = "ana_bgen14a2_sel5z_s6.root"
#
basedir_gg = "../../../star-upc-data/ana/starsim/slight14e/sel5"
infile_gg = "ana_slight14e2x1_sel5_nzvtx.root"
#model predictions
gSlight = load_starlight(dy)
gSartre = load_sartre()
gFlat = loat_flat_pt2()
gMS = load_ms()
gCCK = load_cck()
#open the inputs
inp = TFile.Open(basedir + "/" + infile)
tree = inp.Get("jRecTree")
#
inp_gg = TFile.Open(basedir_gg + "/" + infile_gg)
tree_gg = inp_gg.Get("jRecTree")
#
inp_sl = TFile.Open(basedir_sl + "/" + infile_sl)
tree_sl_gen = inp_sl.Get("jGenTree")
#
inp_sart = TFile.Open(basedir_sart + "/" + infile_sart)
tree_sart_gen = inp_sart.Get("jGenTree")
#
inp_bgen = TFile.Open(basedir_bgen + "/" + infile_bgen)
tree_bgen_gen = inp_bgen.Get("jGenTree")
#evaluate binning
#print "bins:", ut.get_nbins(ptbin, ptmin, ptmax)
bins = ut.get_bins_vec_2pt(ptbin, ptlon, ptmin, ptmax, ptmid)
#bins = ut.get_bins_vec_3pt(ptshort, ptbin, ptlon, ptmin, ptmax, ptlow, ptmid)
#print "bins2:", bins.size()-1
#load the data
strsel = "jRecM>{0:.3f} && jRecM<{1:.3f}".format(mmin, mmax)
hPt = ut.prepare_TH1D_vec("hPt", bins)
tree.Draw("jRecPt*jRecPt >> hPt", strsel)
#distribution for bin centers
hPtCen = hPt.Clone("hPtCen")
#gamma-gamma component
hPtGG = ut.prepare_TH1D_vec("hPtGG", bins)
tree_gg.Draw("jRecPt*jRecPt >> hPtGG", strsel)
#normalize the gamma-gamma component
ut.norm_to_num(hPtGG, ngg, rt.kGreen)
#incoherent functional shape
func_incoh_pt2 = TF1("func_incoh", "[0]*exp(-[1]*x)", 0., 10.)
func_incoh_pt2.SetParameters(873.04, 3.28)
#fill incoherent histogram from functional shape
hPtIncoh = ut.prepare_TH1D_vec("hPtIncoh", bins)
ut.fill_h1_tf(hPtIncoh, func_incoh_pt2, rt.kRed)
#print "Entries before gamma-gamma and incoherent subtraction:", hPt.GetEntries()
#subtract gamma-gamma and incoherent components
hPt.Sumw2()
hPt.Add(hPtGG, -1)
#print "Gamma-gamma entries:", hPtGG.Integral()
#print "Entries after gamma-gamma subtraction:", hPt.Integral()
#print "Incoherent entries:", hPtIncoh.Integral()
hPt.Add(hPtIncoh, -1)
#print "Entries after all subtraction:", hPt.Integral()
#scale the luminosity
lumi_scaled = lumi * ratio_ana * ratio_zdc_vtx
#print "lumi_scaled:", lumi_scaled
#denominator for deconvoluted distribution, conversion ub to mb
den = Reta * br * zdc_acc * trg_eff * bbceff * ratio_tof * lumi_scaled * 1000. * dy
#deconvolution
deconv_min = bins[0]
deconv_max = bins[bins.size() - 1]
deconv_nbin = bins.size() - 1
gROOT.LoadMacro("fill_response_matrix.C")
#Starlight response
#resp_sl = RooUnfoldResponse(deconv_nbin, deconv_min, deconv_max, deconv_nbin, deconv_min, deconv_max)
resp_sl = RooUnfoldResponse(hPt, hPt)
rt.fill_response_matrix(tree_sl_gen, resp_sl)
#
unfold_sl = RooUnfoldBayes(resp_sl, hPt, 15)
#unfold_sl = RooUnfoldSvd(resp_sl, hPt, 15)
hPtSl = unfold_sl.Hreco()
#ut.set_H1D(hPtSl)
#apply the denominator and bin width
ut.norm_to_den_w(hPtSl, den)
#Sartre response
#resp_sart = RooUnfoldResponse(deconv_nbin, deconv_min, deconv_max, deconv_nbin, deconv_min, deconv_max)
#resp_sart = RooUnfoldResponse(hPt, hPt)
#rt.fill_response_matrix(tree_sart_gen, resp_sart)
#
#unfold_sart = RooUnfoldBayes(resp_sart, hPt, 10)
#hPtSart = unfold_sart.Hreco()
#ut.set_H1D(hPtSart)
#hPtSart.SetMarkerStyle(21)
#Flat pT^2 response
#resp_bgen = RooUnfoldResponse(deconv_nbin, deconv_min, deconv_max, deconv_nbin, deconv_min, deconv_max)
resp_bgen = RooUnfoldResponse(hPt, hPt)
rt.fill_response_matrix(tree_bgen_gen, resp_bgen)
#
unfold_bgen = RooUnfoldBayes(resp_bgen, hPt, 14)
hPtFlat = unfold_bgen.Hreco()
#ut.set_H1D(hPtFlat)
#apply the denominator and bin width
ut.norm_to_den_w(hPtFlat, den)
#hPtFlat.SetMarkerStyle(22)
#hPtFlat.SetMarkerSize(1.3)
#systematical errors
err_zdc_acc = 0.1
err_bemc_eff = 0.03
#sys_err = rt.TMath.Sqrt(err_zdc_acc*err_zdc_acc + err_bemc_eff*err_bemc_eff)
sys_err = err_zdc_acc * err_zdc_acc + err_bemc_eff * err_bemc_eff
#print "Total sys err:", sys_err
hSys = ut.prepare_TH1D_vec("hSys", bins)
hSys.SetOption("E2")
hSys.SetFillColor(rt.kOrange + 1)
hSys.SetLineColor(rt.kOrange)
for ibin in xrange(1, hPtFlat.GetNbinsX() + 1):
hSys.SetBinContent(ibin, hPtFlat.GetBinContent(ibin))
sig_sl = hPtSl.GetBinContent(ibin)
sig_fl = hPtFlat.GetBinContent(ibin)
err_deconv = TMath.Abs(sig_fl - sig_sl) / sig_fl
#print "err_deconv", err_deconv
#sys_err += err_deconv*err_deconv
sys_err_sq = sys_err + err_deconv * err_deconv
sys_err_bin = TMath.Sqrt(sys_err_sq)
stat_err = hPtFlat.GetBinError(ibin) / hPtFlat.GetBinContent(ibin)
tot_err = TMath.Sqrt(stat_err * stat_err + sys_err_sq)
#hSys.SetBinError(ibin, hPtFlat.GetBinContent(ibin)*err_deconv)
hSys.SetBinError(ibin, hPtFlat.GetBinContent(ibin) * sys_err_bin)
#hPtFlat.SetBinError(ibin, hPtFlat.GetBinContent(ibin)*tot_err)
#draw the results
gStyle.SetPadTickX(1)
gStyle.SetFrameLineWidth(2)
#frame for models plot only
frame = ut.prepare_TH1D("frame", ptbin, ptmin, ptmax)
can = ut.box_canvas()
#ut.set_margin_lbtr(gPad, 0.1, 0.09, 0.03, 0.03)
ut.set_margin_lbtr(gPad, 0.1, 0.09, 0.055, 0.01)
ytit = "d#it{#sigma}/d#it{t}d#it{y} (mb/(GeV/c)^{2})"
xtit = "|#kern[0.3]{#it{t}}| ((GeV/c)^{2})"
ut.put_yx_tit(frame, ytit, xtit, 1.4, 1.2)
frame.SetMaximum(11)
#frame.SetMinimum(1.e-6)
#frame.SetMinimum(2e-4)
frame.SetMinimum(1e-5) # 3e-5
frame.Draw()
#hSys.Draw("e2same")
#bin center points from data
#gSig = apply_centers(hPtFlat, hPtCen)
gSig = fixed_centers(hPtFlat)
ut.set_graph(gSig)
#hPtSl.Draw("e1same")
#hPtSart.Draw("e1same")
#hPtFlat.Draw("e1same")
#put model predictions
#gSartre.Draw("lsame")
#gFlat.Draw("lsame")
gMS.Draw("lsame")
gCCK.Draw("lsame")
gSlight.Draw("lsame")
gSig.Draw("P")
frame.Draw("same")
gPad.SetLogy()
cleg = ut.prepare_leg(0.1, 0.96, 0.14, 0.01, 0.035)
cleg.AddEntry(
None,
"Au+Au #rightarrow J/#psi + Au+Au + XnXn, #sqrt{#it{s}_{#it{NN}}} = 200 GeV",
"")
cleg.Draw("same")
leg = ut.prepare_leg(0.45, 0.82, 0.18, 0.1, 0.035)
leg.AddEntry(None, "#bf{|#kern[0.3]{#it{y}}| < 1}", "")
hx = ut.prepare_TH1D("hx", 1, 0, 1)
leg.AddEntry(hx, "STAR")
hx.Draw("same")
leg.Draw("same")
#legend for models
mleg = ut.prepare_leg(0.68, 0.76, 0.3, 0.16, 0.035)
#mleg = ut.prepare_leg(0.68, 0.8, 0.3, 0.12, 0.035)
mleg.AddEntry(gSlight, "STARLIGHT", "l")
mleg.AddEntry(gMS, "MS", "l")
mleg.AddEntry(gCCK, "CCK-hs", "l")
#mleg.AddEntry(gSartre, "Sartre", "l")
#mleg.AddEntry(gFlat, "Flat #it{p}_{T}^{2}", "l")
mleg.Draw("same")
#legend for deconvolution method
dleg = ut.prepare_leg(0.3, 0.75, 0.2, 0.18, 0.035)
#dleg = ut.prepare_leg(0.3, 0.83, 0.2, 0.1, 0.035)
dleg.AddEntry(None, "Unfolding with:", "")
dleg.AddEntry(hPtSl, "Starlight", "p")
#dleg.AddEntry(hPtSart, "Sartre", "p")
dleg.AddEntry(hPtFlat, "Flat #it{p}_{T}^{2}", "p")
#dleg.Draw("same")
ut.invert_col(rt.gPad)
can.SaveAs("01fig.pdf")
#to prevent 'pure virtual method called'
gPad.Close()
#save the cross section to output file
out = TFile("sigma.root", "recreate")
gSig.Write("sigma")
out.Close()
#beep when finished
gSystem.Exec("mplayer ../computerbeep_1.mp3 > /dev/null 2>&1")
# get pythia 8 reco and normalize
pdf_reco = pdffile.Get('PFJet_pt_m_AK8')
pdf_reco_softdrop = pdffile.Get('PFJet_pt_m_AK8SD')
pdf_reco.Scale(1. / pdf_reco.Integral())
pdf_reco_softdrop.Scale(1. / pdf_reco_softdrop.Integral())
# get truth and normalize it
pdf_gen = pdffile.Get('PFJet_pt_m_AK8Gen')
pdf_gen_softdrop = pdffile.Get('PFJet_pt_m_AK8SDgen')
pdf_gen.Scale(1. / pdf_gen.Integral())
pdf_gen_softdrop.Scale(1. / pdf_gen_softdrop.Integral())
##################################################################################################### Unfold Pythia8 with PDF-UP
unfold_pdfup = RooUnfoldBayes(pdfup_response, pdf_reco, 4)
unfolded_pdfup = unfold_pdfup.Hreco()
canvases_up = []
namesreco_up = []
legends_up = []
for x in range(0, nptbin):
canvases_up.append(TCanvas("canvas_pdfup" + str(x)))
namesreco_up.append(None)
legends_up.append(TLegend(.7, .5, .9, .7))
for i, canvas in enumerate(canvases_up):
canvas.cd()
namesreco_up[i] = unfolded_pdfup.ProjectionX('pdf_up' + str(i), i + 1,
i + 1)
# hDiff_tmp = TH1F("diff_bin"+str(ibin)+"_"+str(iiter),"; unfolded-truth; number of events",100,-0.2,0.2)
#elif ibin < 7:
# hDiff_tmp = TH1F("diff_bin"+str(ibin)+"_"+str(iiter),"; unfolded-truth; number of events",100,-0.4,0.4)
#else :
hDiff_tmp = TH1F("diff_bin"+str(ibin)+"_"+str(iiter),"; unfolded-truth; number of events",100,-1,1)
hDiff_bin.append(hDiff_tmp)
lowedge = 399.
highedge = 1199.
for iiter in xrange(1,tot_iter+1) :
for itoy in xrange(0,ntoys) :
#print 'iiter = ', iiter, ' itoy = ', itoy
unfold = RooUnfoldBayes(response, hToy_i[itoy], iiter)
#unfold = RooUnfoldSvd(response, hToy_i[itoy], iiter)
hReco_tmp = unfold.Hreco()
for ibin in range(0, nbins):
if myTrue.GetBinLowEdge(ibin+1) > lowedge and myTrue.GetBinLowEdge(ibin+1) < highedge:
hDiff_bin[(iiter-1)*nbins + ibin].Fill( (myTrue.GetBinContent(ibin+1) - hReco_tmp.GetBinContent(ibin+1)) / myTrue.GetBinContent(ibin+1))
color = [1,2,3,4,5,6,7,8,9,14]
for ibin in xrange(0,nbins) :
c = TCanvas()
if hDiff_bin[1*nbins + ibin].GetSum() == 0:
continue
reco.Scale(1. / reco.Integral())
truthSD.Scale(1. / truthSD.Integral())
recoSD.Scale(1. / recoSD.Integral())
pt_bin = {
0: '200-240',
1: '240-310',
2: '310-400',
3: '400-530',
4: '530-650',
5: '650-760',
6: '760-Inf'
}
unfold = RooUnfoldBayes(response, reco, 6)
unfoldSD = RooUnfoldBayes(responseSD, recoSD, 6)
#unfold= RooUnfoldSvd(response, reco, 5);
reco_unfolded = unfold.Hreco()
reco_unfoldedSD = unfoldSD.Hreco()
################### New Correlation matrix stuff
cov = unfold.Ereco()
covSD = unfoldSD.Ereco()
nb = cov.GetNrows()
import math
cor = ROOT.TH2F("cor", "", nb, 0, nb, nb, 0, nb)
corSD = ROOT.TH2F("corSD", "", nb, 0, nb, nb, 0, nb)
herwig_reco.Scale(1. / herwig_reco.Integral())
herwig_reco_softdrop.Scale(1. / herwig_reco_softdrop.Integral())
# get truth and normalize it
pythia8_gen = pythia8file.Get('PFJet_pt_m_AK8Gen')
pythia8_gen_softdrop = pythia8file.Get('PFJet_pt_m_AK8SDgen')
pythia8_gen.Scale(1. / pythia8_gen.Integral())
pythia8_gen_softdrop.Scale(1. / pythia8_gen_softdrop.Integral())
herwig_gen = herwigfile.Get('PFJet_pt_m_AK8Gen')
herwig_gen_softdrop = herwigfile.Get('PFJet_pt_m_AK8SDgen')
herwig_gen.Scale(1. / herwig_gen.Integral())
herwig_gen_softdrop.Scale(1. / herwig_gen_softdrop.Integral())
########################################################################################################### Unfold ungroomed pythia 8 with herwig
unfold_ps = RooUnfoldBayes(herwig_response, pythia8_reco, 4)
unfolded_ps = unfold_ps.Hreco()
pyth8_uherwig = unfolded_ps.ProjectionX()
pyth8_uherwig.Scale(1.0 / pyth8_uherwig.Integral())
pyth8_uherwig.SetName("pyth8_uherwig")
canvases = []
namesreco = []
namesgen = []
legends = []
for x in range(0, nptbin):
canvases.append(TCanvas("canvas" + str(x)))
namesreco.append(None)
namesgen.append(None)
legends.append(TLegend(.7, .5, .9, .7))
for i, canvas in enumerate(canvases):
elif xt == xt and xm != xm:
response.Miss(xt)
for i in range(N / 2, N):
tree.GetEntry(i)
xm = tree._recoTop_0_Eta
xt = tree.trueTop_genParticleSelector_0_Eta
if xm == xm:
hMeas_test.Fill(xm)
if xt == xt:
hTrue_test.Fill(xt)
of = ROOT.TFile("out_unfold.root", "RECREATE" if unfold else "READ")
if unfold:
unfoldBayes = RooUnfoldBayes(response, hMeas_test, 4, False, "bayes")
kReg = 10 #Nbins/2
unfoldSVD = RooUnfoldSvd(response, hMeas_test, kReg, 1000, "svd")
unfoldTUnf = RooUnfoldTUnfold(response, hMeas_test,
ROOT.TUnfold.kRegModeDerivative, "tunf")
hRecoBayes = unfoldBayes.Hreco()
hRecoSVD = unfoldSVD.Hreco()
hRecoTUnf = unfoldTUnf.Hreco()
of.Write()
else:
hRecoBayes = of.Get("bayes")
hRecoSVD = of.Get("svd")
hRecoTUnf = of.Get("tunf")
from ROOT import gRandom, TH1, cout, TH2, TLegend, TFile
from ROOT import RooUnfoldResponse
from ROOT import RooUnfold
from ROOT import RooUnfoldBayes
from ROOT import TCanvas
from ROOT import RooUnfoldSvd
myfile = TFile('qcd_2d_unfold_MC.root')
response = myfile.Get('2d_response')
truth = myfile.Get('PFJet_pt_m_AK8Gen')
reco = myfile.Get('PFJet_pt_m_AK8')
response.Draw('colz')
outfile = TFile('unfolded2d.root', 'RECREATE')
unfold = RooUnfoldBayes(response, reco, 6)
#unfold= RooUnfoldSvd(response, reco, 5);
reco_unfolded = unfold.Hreco()
reco_unfolded.Draw()
truth.SetLineColor(4)
truth.Draw('SAME')
c2 = TCanvas()
c3 = TCanvas()
c4 = TCanvas()
c5 = TCanvas()
c6 = TCanvas()
def main(optunf="Bayes"):
optunfs = ["Bayes", "SVD", "TUnfold", "Invert", "Reverse"]
if not optunf in optunfs:
txt = "Unfolding option " + optunf + " not recognised"
raise ValueError(txt)
global hReco, hMeas, hTrue
print "==================================== TRAIN ===================================="
# Create response matrix object for 40 measured and 20
# unfolded bins:
response = RooUnfoldResponse(40, -10.0, 10.0, 20, -10.0, 10.0)
# Train with a Breit-Wigner, mean 0.3 and width 2.5.
for i in xrange(100000):
# xt= gRandom.BreitWigner( 0.3, 2.5 )
xt = gRandom.Gaus(0.0, 5.0)
x = smear(xt)
if x != None:
response.Fill(x, xt)
else:
response.Miss(xt)
print "==================================== TEST ====================================="
hTrue = TH1D("true", "Test Truth", 20, -10.0, 10.0)
hMeas = TH1D("meas", "Test Measured", 40, -10.0, 10.0)
# Test with a Gaussian, mean 0 and width 2.
for i in xrange(10000):
# xt= gRandom.Gaus( 0.0, 2.0 )
xt = gRandom.BreitWigner(0.3, 2.5)
x = smear(xt)
hTrue.Fill(xt)
if x != None:
hMeas.Fill(x)
print "==================================== UNFOLD ==================================="
print "Unfolding method:", optunf
if "Bayes" in optunf:
# Bayes unfoldung with 4 iterations
# unfold= RooUnfoldBayes( response, hMeas, 4 )
unfold = RooUnfoldBayes(response, hMeas, 10, False, True)
elif "SVD" in optunf:
# SVD unfoding with free regularisation
# unfold= RooUnfoldSvd( response, hMeas, 20 )
unfold = RooUnfoldSvd(response, hMeas)
elif "TUnfold" in optunf:
# TUnfold with fixed regularisation tau=0.002
# unfold= RooUnfoldTUnfold( response, hMeas )
unfold = RooUnfoldTUnfold(response, hMeas, 0.002)
elif "Invert" in optunf:
unfold = RooUnfoldInvert(response, hMeas)
elif "Reverse" in optunf:
unfold = RooUnfoldBayes(response, hMeas, 1)
hReco = unfold.Hreco()
# unfold.PrintTable( cout, hTrue )
unfold.PrintTable(cout, hTrue, 2)
hReco.Draw()
hMeas.Draw("SAME")
hTrue.SetLineColor(8)
hTrue.Draw("SAME")
return
matrix_unfolded1 = np.zeros((Nbins,Nbins))
for i_Eg1 in range(Nbins):#range(60,62):
hMeas= TH1D ("meas", "Test Measured", Nbins, Emin, Emax);
print("Now doing i_Eg1 =", i_Eg1, flush=True)
for i in range(Nbins):
Ei = E_resp_array[i]
hMeas.Fill(Ei,matrix_folded[i_Eg1,i])
# hack to recalculate the Uncertainties now, after the histogram is filled
hMeas.Sumw2(False)
hMeas.Sumw2(True)
# hTrue.Sumw2(False) # doesn't work yet?
# hTrue.Sumw2(True) # doesn't work yet?
# print("==================================== UNFOLD ===================================")
unfold= RooUnfoldBayes (response, hMeas, Niterations); # OR
# unfold= RooUnfoldSvd (response, hMeas, 20); # OR
#unfold= RooUnfoldTUnfold (response, hMeas); # OR
# unfold= RooUnfoldIds (response, hMeas, 3); # OR
# unfold= RooUnfoldInvert (response, hMeas); # OR
hReco= unfold.Hreco();
# unfold.PrintTable (cout, hTrue);
matrix_unfolded1[i_Eg1,:] = np.array(hReco)[0:Nbins]
matrix_unfolded1 = np.nan_to_num(matrix_unfolded1)
np.save(fname_save_unf1, matrix_unfolded1)
h_data_toUse.SetBinContent(bin, newContent)
h_data_toUse.SetBinError(bin, sqrt(newContent))
h_measured_toUse = asrootpy(h_measured)
# Remove fakes before unfolding
h_fakes = h_measured_toUse - h_response.ProjectionX()
nonFakeRatio = 1 - h_fakes / h_measured_toUse
h_measured_toUse *= nonFakeRatio / nonFakeRatio
h_data_toUse *= nonFakeRatio / nonFakeRatio
# Unfold with SVD
response = RooUnfoldResponse(h_measured_toUse, h_truth, h_response)
svdUnfolding = RooUnfoldSvd(response, h_data_toUse, 0.7)
svdUnfolding.SetVerbose(0)
h_unfolded_data_svd = svdUnfolding.Hreco(3)
# Unfold with Bayes
response_bayes = RooUnfoldResponse(h_measured_toUse, h_truth, h_response)
bayesUnfolding = RooUnfoldBayes(response_bayes, h_data_toUse, 4)
h_unfolded_data_bayes = bayesUnfolding.Hreco(3)
# Store result of first bin
h_svdFirstBin.Fill(h_unfolded_data_svd.GetBinContent(1))
h_bayesFirstBin.Fill(h_unfolded_data_bayes.GetBinContent(1))
# Draw histograms
h_svdFirstBin.Draw()
raw_input('SVD')
h_bayesFirstBin.Draw()
raw_input('Bayes')
histBE = checkBE.Hreco()
histBE.SetLineColor(2)
c4 = ROOT.TCanvas("c4", "c4", 800, 800)
c4.SetLogx()
c4.SetLogy()
histBE.GetXaxis().SetTitle("m[GeV]")
histBE.SetTitle("Unfolded reco and gen for BE")
DY2 = ROOT.TLegend(0.1, 0.7, 0.3, 0.9)
DY2.AddEntry(histBE, "reco")
DY2.AddEntry(genBE, "gen")
histBE.Draw("hist")
genBE.Draw("samehist")
DY2.Draw()
c4.Print("test4.pdf")
checkBB1 = RooUnfoldBayes(responseBB, recBB)
histBB1 = checkBB1.Hreco()
histBB1.SetLineColor(2)
c5 = ROOT.TCanvas("c5", "c5", 800, 800)
c5.SetLogx()
c5.SetLogy()
histBB1.GetXaxis().SetTitle("m[GeV]")
histBB1.SetTitle("Unfolded reco and gen for BB")
DY3 = ROOT.TLegend(0.1, 0.7, 0.3, 0.9)
DY3.AddEntry(histBB1, "reco")
DY3.AddEntry(genBB, "gen")
histBB1.Draw("hist")
genBB.Draw("samehist")
DY3.Draw()
c5.Print("test5.pdf")
def doUnfold(self, data, back=0, regdiff=0, iteration=0):
self.data = data.Clone()
if back != 0:
self.data.Sumw2()
back.Sumw2()
self.data.Add(back, -1.)
self.back = back.Clone()
for i in range(self.data.GetNbinsX() + 2):
if self.data.GetBinContent(i) < 0:
self.data.SetBinContent(i, 0)
if regdiff != 0:
Unfold = RooUnfoldBayes(self.response, self.data, 1, False,
self.name + "_reg", self.name + "_reg")
h2 = self.data.Clone()
for i in range(9999):
h1 = Unfold.Hreco().Clone()
vaild = 0
val = 0
for j in range(1, h2.GetNbinsX() + 1):
if h2.GetBinContent(j) > 0:
val = h1.GetBinContent(j) / h2.GetBinContent(j)
val = abs(val - 1)
if vaild < val:
vaild = val
print vaild
if vaild < regdiff:
print "%d times iteration" % i
self.UnfoldedData = h1
return h1
else:
h2 = h1.Clone()
Unfold.Reset()
Unfold = RooUnfoldBayes(self.response, self.data, i + 2,
False, self.name + "_reg",
self.name + "_reg")
elif iteration != 0:
s = "iter_%d" % iteration
Unfold = RooUnfoldBayes(self.response, self.data, iteration, False,
self.name + s, self.name + s)
h1 = Unfold.Hreco().Clone()
return h1
else:
Unfold = RooUnfoldInvert(self.response, self.data,
self.name + "_invert",
self.name + "_invert")
Unfold.SetMeasured(self.data)
self.Unfold = Unfold
h1 = Unfold.Hreco().Clone()
self.UnfoldedData = h1
return h1
matrix_unfolded1 = np.zeros((Nbins, Nbins))
for i_Eg1 in range(Nbins): #range(60,62):
hMeas = TH1D("meas", "Test Measured", Nbins, Emin, Emax)
print("Now doing i_Eg1 =", i_Eg1, flush=True)
for i in range(Nbins):
Ei = E_resp_array[i]
hMeas.Fill(Ei, matrix_folded[i_Eg1, i])
# hack to recalculate the Uncertainties now, after the histogram is filled
hMeas.Sumw2(False)
hMeas.Sumw2(True)
# hTrue.Sumw2(False) # doesn't work yet?
# hTrue.Sumw2(True) # doesn't work yet?
# print("==================================== UNFOLD ===================================")
unfold = RooUnfoldBayes(response, hMeas, Niterations)
# OR
# unfold= RooUnfoldSvd (response, hMeas, 20); # OR
#unfold= RooUnfoldTUnfold (response, hMeas); # OR
# unfold= RooUnfoldIds (response, hMeas, 3); # OR
# unfold= RooUnfoldInvert (response, hMeas); # OR
hReco = unfold.Hreco()
# unfold.PrintTable (cout, hTrue);
matrix_unfolded1[i_Eg1, :] = np.array(hReco)[0:Nbins]
matrix_unfolded1 = np.nan_to_num(matrix_unfolded1)
np.save(fname_save_unf1, matrix_unfolded1)
# cbar_ax3 = ax3.pcolormesh(E_array_folded, E_array_folded, matrix_unfolded1, norm=LogNorm(vmin=1e-1, vmax=1e3))
responseSD = mcfile.Get('2d_response_softdrop' + options.extension)
truth = mcfile.Get('PFJet_pt_m_AK8Gen')
truthSD = mcfile.Get('PFJet_pt_m_AK8SDgen')
reco = datafile.Get('PFJet_pt_m_AK8')
recoSD = datafile.Get('PFJet_pt_m_AK8SD')
truth.Scale(1. / truth.Integral())
reco.Scale(1. / reco.Integral())
truthSD.Scale(1. / truthSD.Integral())
recoSD.Scale(1. / recoSD.Integral())
response.Draw('colz')
unfold = RooUnfoldBayes(response, reco, 4)
unfoldSD = RooUnfoldBayes(responseSD, recoSD, 4)
reco_unfolded = unfold.Hreco()
recoSD_unfolded = unfoldSD.Hreco()
reco_unfolded.Draw()
truth.SetLineColor(4)
truth.Draw('SAME')
pt_bin = {
0: '200 < p_{T} < 260',
1: '260 < p_{T} < 350',
2: '350 < p_{T} < 460',
DnResponse = g.Get('2d_response_jecdn')
UpResponseSD = g.Get('2d_response_softdrop_jecup')
DnResponseSD = g.Get('2d_response_softdrop_jecdn')
else:
UpResponse = g.Get('2d_response_' + options.sys + 'up')
DnResponse = g.Get('2d_response_' + options.sys + 'dn')
UpResponseSD = g.Get('2d_response_softdrop_' + options.sys + 'up')
DnResponseSD = g.Get('2d_response_softdrop_' + options.sys + 'dn')
if options.hist == 'h_msd':
unfold_hnom = RooUnfoldBayes(NomResponseSD, hnom1, 4)
unfold_hup = RooUnfoldBayes(UpResponseSD, hnom1, 4)
unfold_hdn = RooUnfoldBayes(DnResponseSD, hnom1, 4)
else:
unfold_hnom = RooUnfoldBayes(NomResponse, hnom1, 4)
unfold_hup = RooUnfoldBayes(UpResponse, hnom1, 4)
unfold_hdn = RooUnfoldBayes(DnResponse, hnom1, 4)
hnom2 = unfold_hnom.Hreco()
hup2 = unfold_hup.Hreco()
hdn2 = unfold_hdn.Hreco()
hnom = hnom2.ProjectionY()
hup = hup2.ProjectionY()
fout = TFile(
"UnfoldingPlots/unfold_" + options.toUnfold + "_PowhegPythia8_" +
options.lepType + "_" + syst_flag + closureout + norm_flag + ".root",
"recreate")
# -------------------------------------------------------------------------------------
# do the actual unfolding
# -------------------------------------------------------------------------------------
print "------------ UNFOLDING (syst: " + syst_flag + ") ------------"
n_iter = 3
print " using " + str(n_iter) + " iterations"
unfold = RooUnfoldBayes(response, hMeas, n_iter)
#unfold = RooUnfoldSvd(response, hMeas, 2);
# get the unfolded distribution
hReco = unfold.Hreco()
hReco.Sumw2()
# -------------------------------------------------------------------------------------
# Translate to cross section (not events) in bins of pt N/L/BR)
# -------------------------------------------------------------------------------------
# TODO: should fix BR
print "WARNING: treatment of branching ratio is wrong!"
hTrue.Scale(1.0 / (lum * 0.438 / 3.)) # true @ parton level
hMeas.Scale(1.0 / (lum * 0.438 / 3.)) # measured @ reco level
