def buildCorrectionHisto(
hndRealDataCr,
hndFakeDataLo,
hndFakeDataHi,
hndFakeMcLo,
hndFakeMcHi,
histoname="lep_corHFRate",
nIter=1,
verbose=False,
plotdir=None,
):
hRealEff = buildRatioHistogram(hndRealDataCr["num"], hndRealDataCr["den"], "real_eff")
corrected = dict([(nd, hndFakeDataLo[nd].Clone("corrected_" + nd)) for nd in ["num", "den"]])
print "buildCorrectionHisto with nIter ", nIter
can = r.TCanvas("c_" + histoname, "")
can.Draw()
can.cd()
hRealEff.GetYaxis().SetRangeUser(0.0, 1.0)
hRealEff.Draw("axis")
tex = r.TLatex()
tex.SetNDC(True)
tex.DrawLatex(0.15, 0.915, histoname)
can._tex = tex
can._histos = [hRealEff]
can._leg = r.TLegend(0.925, 0.25, 1.0, 0.9, "iter")
can._leg.SetBorderSize(0)
can._leg.SetFillColor(0)
for iteration in range(nIter):
print "iter ", iteration
rate = buildRatioHistogram(corrected["num"], corrected["den"]) # temporary rate (?)
if verbose:
print "Iteration %d, corrected values:" % iteration
print " num %s" % lf2s(getBinContents(corrected["num"]))
print " den %s" % lf2s(getBinContents(corrected["den"]))
print " ratio %s" % lf2s(getBinContents(rate))
dataNum, dataDen = hndFakeDataHi["num"], hndFakeDataHi["den"]
for nd, tl in [("num", "tight"), ("den", "loose")]:
corr, dataLow = corrected[nd], hndFakeDataLo[nd]
mcLow, mcHi = hndFakeMcLo[nd], hndFakeMcHi[nd]
corrFact = getCorrFactors(hRealEff, rate, dataNum, dataDen, mcHi, tl)
corr = correctRate(corr, dataLow, mcLow, corrFact)
rate.SetLineColor(20 + iteration)
rate.SetMarkerColor(20 + iteration)
rate.SetLineWidth(4)
rate.SetMarkerStyle(r.kFullCross)
can._histos.append(rate.DrawClone("el same"))
can._leg.AddEntry(can._histos[-1], "%d" % iteration, "lp")
ratio = buildRatioHistogram(corrected["num"], corrected["den"], histoname)
can._leg.Draw()
can.Update()
if plotdir:
can.SaveAs(plotdir + "/" + histoname + "_iterations.png")
return ratio
def computeAndPlotHfSf2d(fileIter, fileHf, lepton, variable_name, outdir) :
eff_da = fileIter.Get(lepton+'_corHFRate_eta')
eff_mc = buildRate(fileHf, lepton+'_fakeHF_all_l_pt_eta')
ratio = buildRatioHistogram(eff_da, eff_mc)
ratio.Print()
xAx, yAx = ratio.GetXaxis(), ratio.GetYaxis()
print ratio.GetName(),": bins (%d, %d)"%(ratio.GetNbinsX(), ratio.GetNbinsY())
nEtaBins = yAx.GetNbins()
print 'nEtaBins: ',nEtaBins
xMin, xMax = xAx.GetXmin(), xAx.GetXmax()
fitFunc = r.TF1('fit_func_const_'+ratio.GetName(), '[0]', xMin, xMax)
etaBins = range(1, 1+nEtaBins)
slices = [ratio.ProjectionX("%s_bin%d"%(ratio.GetName(), b), b, b, 'e') for b in etaBins]
for b, s in zip(etaBins, slices) :
s.SetTitle(lepton+" data/mc heavyflavor : eta bin %d"%b)
s.Fit(fitFunc.GetName(), '0RQ') # do not draw, range, quiet
p0, p0Err, chi2, ndf = fitResults(fitFunc)
p0, p0Err = pdgRound(p0, p0Err)
print "bin %d : %s +/- %s"%(b, p0, p0Err)
can = r.TCanvas('')
s.Draw('ep')
fitFunc.Draw('same')
tex = r.TLatex()
tex.SetNDC(True)
fitParLabel = "Const. fit : %s #pm %s"%(p0, p0Err)
fitGoodLabel = "#chi^{2}/DOF : %.2f / %d"%(chi2, ndf)
tex.SetTextSize(yAx.GetTitleSize())
tex.SetTextFont(yAx.GetTitleFont())
tex.DrawLatex(0.15, 0.45, s.GetTitle())
tex.DrawLatex(0.15, 0.40, "#splitline{%s}{%s}"%(fitParLabel, fitGoodLabel))
can.Update()
for ext in ['eps','png'] : can.SaveAs(outdir+'/fit_'+lepton+"_heavyflavor_etabin%d.%s"%(b, ext))
return p0
def buildSideBandSubRate(file, lepton, variable_name):
hZwindow = getNumDenHistos(file, lepton + '_realCR_' + variable_name)
hSideLo = getNumDenHistos(file, lepton + '_realSideLow_' + variable_name)
hSideHi = getNumDenHistos(file, lepton + '_realSideHigh_' + variable_name)
def sbErr(el, ew, eh):
"SidebandError; DG this doesn't make any sense to me; ask Matt"
return sqrt(ew * ew - (el * el + eh * eh))
def be(h):
return [h.GetBinError(b) for b in range(1, 1 + h.GetNbinsX())]
errs = dict([(k, [
sbErr(l, w, h)
for l, w, h in zip(be(hSideLo[k]), be(hZwindow[k]), be(hSideHi[k]))
]) for k in ['num', 'den']])
hSideLo['num'].Add(hSideHi['num'])
hSideLo['den'].Add(hSideHi['den'])
hZwindow['num'].Add(hSideLo['num'], -1)
hZwindow['den'].Add(hSideLo['den'], -1)
for nd in ['num', 'den']:
h, err = hZwindow[nd], errs[nd]
nbins = h.GetNbinsX()
for i, b in zip(range(nbins), range(1, 1 + nbins)):
h.SetBinError(b, err[i])
return buildRatioHistogram(hZwindow['num'], hZwindow['den'])
def computeAndPlotConvSf(fileData, fileMc, lepton, variable_name, outdir):
"Electron conversion: simplest case, just data/mc"
eff_da = buildRate(fileData, lepton + '_fakeConv_' + variable_name)
eff_mc = buildRate(fileMc, lepton + '_fakeConv_' + variable_name)
ratio = buildRatioHistogram(eff_da, eff_mc)
fitFunc = fitWithConst(ratio)
p0, p0Err, chi2, ndf = fitResults(fitFunc)
p0, p0Err = pdgRound(p0, p0Err)
print "SF for %s fake conv : %s +/- %s" % (lepton, p0, p0Err)
graphics = {
'xtitle': xTitle(lepton, variable_name),
'ytitle': lepton + ' p(tight | fake conv)',
'colors': {
'data': r.kBlack,
'mc': mcColor(lepton)
},
'markers': {
'data': r.kFullCircle,
'mc': mcMarker(lepton)
},
'labels': {
'data': 'Data: Conversion CR',
'mc': 'MC Comb: Conv CR'
}
}
plotHistRatioAndFit({
'data': eff_da,
'mc': eff_mc
}, ratio, fitFunc, outdir + lepton + '_fakeconv', graphics)
def computeAndPlotRealSf(file_data, file_mc, lepton, variable_name, outdir):
"Scale factor from the real control region, Z tag and probe"
eff_da = buildSideBandSubRate(file_data, lepton, variable_name)
eff_mc = buildSideBandSubRate(file_mc, lepton, variable_name)
ratio = buildRatioHistogram(eff_da, eff_mc)
fitFunc = fitWithConst(ratio)
p0, p0Err, chi2, ndf = fitResults(fitFunc)
p0, p0Err = pdgRound(p0, p0Err)
print "SF for %s real : %s +/- %s" % (lepton, p0, p0Err)
graphics = {
'xtitle': xTitle(lepton, variable_name),
'ytitle': lepton + ' p(tight | real)',
'colors': {
'data': r.kBlack,
'mc': mcColor(lepton)
},
'markers': {
'data': r.kFullCircle,
'mc': mcMarker(lepton)
},
'labels': {
'data': 'Data: Z Tag and Probe',
'mc': 'MC Comb: Z Tag and Probe'
}
}
plotHistRatioAndFit({
'data': eff_da,
'mc': eff_mc
}, ratio, fitFunc, outdir + lepton + '_real', graphics)
def drawBot(pad, histo_data, histo_mc, err_band_r, xaxis_title='') :
pad.Draw()
pad.cd()
ratio = buildRatioHistogram(histo_data, histo_mc)
yMin, yMax = 0.0, 2.0
ratio.SetMinimum(yMin)
ratio.SetMaximum(yMax)
ratio.SetStats(0)
ratio.Draw('axis')
x_lo, x_hi = getXrange(ratio)
refLines = [referenceLine(x_lo, x_hi, y, y) for y in [0.5, 1.0, 1.5]]
for l in refLines : l.Draw()
err_band_r.Draw('E2 same')
ratio.Draw('ep same')
xA, yA = ratio.GetXaxis(), ratio.GetYaxis()
textScaleUp = 1.0/pad.GetHNDC()
if xaxis_title : xA.SetTitle(xaxis_title)
yA.SetNdivisions(-104)
yA.SetTitle('Data/SM')
yA.CenterTitle()
yA.SetTitleOffset(yA.GetTitleOffset()/textScaleUp)
for a in [xA, yA] :
a.SetLabelSize(a.GetLabelSize()*textScaleUp)
a.SetTitleSize(a.GetTitleSize()*textScaleUp)
pad._graphical_objects = [ratio, err_band_r] + refLines # avoid garbage collection
pad.Update()
def computeAndPlotHfSf(fileIter, fileHf, lepton, variable_name, outdir):
"HF tag and probe; in this case we need to subract out the contamination"
eff_da = fileIter.Get(lepton + '_corHFRate')
eff_mc = buildRate(fileHf, lepton + '_fakeHF_' + variable_name)
ratio = buildRatioHistogram(eff_da, eff_mc)
fitFunc = fitWithConst(ratio)
p0, p0Err, chi2, ndf = fitResults(fitFunc)
p0, p0Err = pdgRound(p0, p0Err)
print "SF for %s fake HF : %s +/- %s" % (lepton, p0, p0Err)
graphics = {
'xtitle': xTitle(lepton, variable_name),
'ytitle': lepton + ' p(tight | fake hf)',
'colors': {
'data': r.kBlack,
'mc': mcColor(lepton)
},
'markers': {
'data': r.kFullCircle,
'mc': mcMarker(lepton)
},
'labels': {
'data': 'Data HF Tag and Probe (Iterative Subtraction)',
'mc': 'b#bar{b}/c#bar{c} MC: HF Tag and Probe'
}
}
plotHistRatioAndFit({
'data': eff_da,
'mc': eff_mc
}, ratio, fitFunc, outdir + lepton + '_fakehf', graphics)
def drawBot(pad, histo_data, histo_mc, err_band_r, xaxis_title='') :
pad.Draw()
pad.cd()
ratio = buildRatioHistogram(histo_data, histo_mc)
yMin, yMax = 0.0, 2.0
ratio.SetMinimum(yMin)
ratio.SetMaximum(yMax)
ratio.SetStats(0)
ratio.Draw('axis')
x_lo, x_hi = getXrange(ratio)
refLines = [referenceLine(x_lo, x_hi, y, y) for y in [0.5, 1.0, 1.5]]
for l in refLines : l.Draw()
err_band_r.Draw('E2 same')
ratio.Draw('ep same')
xA, yA = ratio.GetXaxis(), ratio.GetYaxis()
textScaleUp = 1.0/pad.GetHNDC()
if xaxis_title : xA.SetTitle(xaxis_title)
yA.SetNdivisions(-104)
yA.SetTitle('Data/SM')
yA.CenterTitle()
yA.SetTitleOffset(yA.GetTitleOffset()/textScaleUp)
for a in [xA, yA] :
a.SetLabelSize(a.GetLabelSize()*textScaleUp)
a.SetTitleSize(a.GetTitleSize()*textScaleUp)
pad._graphical_objects = [ratio, err_band_r] + refLines # avoid garbage collection
pad.Update()
def computeAndPlotRealSf(file_data, file_mc, lepton, variable_name, outdir) :
"Scale factor from the real control region, Z tag and probe"
eff_da = buildSideBandSubRate(file_data, lepton, variable_name)
eff_mc = buildSideBandSubRate(file_mc, lepton, variable_name)
ratio = buildRatioHistogram(eff_da, eff_mc)
fitFunc = fitWithConst(ratio)
p0, p0Err, chi2, ndf = fitResults(fitFunc)
p0, p0Err = pdgRound(p0, p0Err)
print "SF for %s real : %s +/- %s"%(lepton, p0, p0Err)
graphics = {'xtitle' : xTitle(lepton, variable_name),
'ytitle' : lepton+' p(tight | real)',
'colors' : {'data' : r.kBlack, 'mc' : mcColor(lepton)},
'markers': {'data' : r.kFullCircle, 'mc' : mcMarker(lepton)},
'labels' : {'data' : 'Data: Z Tag and Probe',
'mc' : 'MC Comb: Z Tag and Probe'}}
plotHistRatioAndFit({'data':eff_da, 'mc':eff_mc}, ratio, fitFunc, outdir+lepton+'_real',
graphics)
def computeAndPlotConvSf(fileData, fileMc, lepton, variable_name, outdir) :
"Electron conversion: simplest case, just data/mc"
eff_da = buildRate(fileData, lepton+'_fakeConv_'+variable_name)
eff_mc = buildRate(fileMc, lepton+'_fakeConv_'+variable_name)
ratio = buildRatioHistogram(eff_da, eff_mc)
fitFunc = fitWithConst(ratio)
p0, p0Err, chi2, ndf = fitResults(fitFunc)
p0, p0Err = pdgRound(p0, p0Err)
print "SF for %s fake conv : %s +/- %s"%(lepton, p0, p0Err)
graphics = {'xtitle' : xTitle(lepton, variable_name),
'ytitle' : lepton+' p(tight | fake conv)',
'colors' : {'data' : r.kBlack, 'mc' : mcColor(lepton)},
'markers': {'data' : r.kFullCircle, 'mc' : mcMarker(lepton)},
'labels' : {'data' : 'Data: Conversion CR',
'mc' : 'MC Comb: Conv CR'}}
plotHistRatioAndFit({'data':eff_da, 'mc':eff_mc}, ratio, fitFunc, outdir+lepton+'_fakeconv',
graphics)
def computeAndPlotHfSf(fileIter, fileHf, lepton, variable_name, outdir) :
"HF tag and probe; in this case we need to subract out the contamination"
eff_da = fileIter.Get(lepton+'_corHFRate')
eff_mc = buildRate(fileHf, lepton+'_fakeHF_'+variable_name)
ratio = buildRatioHistogram(eff_da, eff_mc)
fitFunc = fitWithConst(ratio)
p0, p0Err, chi2, ndf = fitResults(fitFunc)
p0, p0Err = pdgRound(p0, p0Err)
print "SF for %s fake HF : %s +/- %s"%(lepton, p0, p0Err)
graphics = {'xtitle' : xTitle(lepton, variable_name),
'ytitle' : lepton+' p(tight | fake hf)',
'colors' : {'data' : r.kBlack, 'mc' : mcColor(lepton)},
'markers': {'data' : r.kFullCircle, 'mc' : mcMarker(lepton)},
'labels' : {'data' : 'Data HF Tag and Probe (Iterative Subtraction)',
'mc' : 'b#bar{b}/c#bar{c} MC: HF Tag and Probe'}}
plotHistRatioAndFit({'data':eff_da, 'mc':eff_mc}, ratio, fitFunc, outdir+lepton+'_fakehf',
graphics)
def computeAndPlotConvSf2d(fileData, fileMc, lepton, variable_name, outdir) :
"Electron conversion: simplest case, just data/mc"
histoname = 'elec_fakeConv_all_l_pt_eta'
eff_da = buildRate(fileData, histoname)
eff_mc = buildRate(fileMc, histoname)
print 'SF conversion: using histo ',histoname
ratio = buildRatioHistogram(eff_da, eff_mc)
ratio.Print()
xAx, yAx = ratio.GetXaxis(), ratio.GetYaxis()
#pt_eta : check that x is pt, y is eta
nEtaBins = yAx.GetNbins()
xMin, xMax = xAx.GetXmin(), xAx.GetXmax()
fitFunc = r.TF1('fit_func_const_'+ratio.GetName(), '[0]', xMin, xMax)
etaBins = range(1, 1+nEtaBins)
slices = [ratio.ProjectionX("%s_bin%d"%(ratio.GetName(), b), b, b, 'e') for b in etaBins]
for b, s in zip(etaBins, slices) :
s.SetTitle("data/mc conversion : eta bin %d"%b)
s.Fit(fitFunc.GetName(), '0RQ') # do not draw, range, quiet
p0, p0Err, chi2, ndf = fitResults(fitFunc)
p0, p0Err = pdgRound(p0, p0Err)
print "bin %d : %s +/- %s"%(b, p0, p0Err)
can = r.TCanvas('')
s.Draw('ep')
fitFunc.Draw('same')
tex = r.TLatex()
tex.SetNDC(True)
fitParLabel = "Const. fit : %s #pm %s"%(p0, p0Err)
fitGoodLabel = "#chi^{2}/DOF : %.2f / %d"%(chi2, ndf)
tex.SetTextSize(yAx.GetTitleSize())
tex.SetTextFont(yAx.GetTitleFont())
tex.DrawLatex(0.15, 0.45, s.GetTitle())
tex.DrawLatex(0.15, 0.40, "#splitline{%s}{%s}"%(fitParLabel, fitGoodLabel))
can.Update()
for ext in ['eps','png'] : can.SaveAs(outdir+"/fit_el_conv_etabin%d.%s"%(b, ext))
# graphics = {'xtitle' : xTitle(lepton, variable_name),
# 'ytitle' : lepton+' p(tight | fake conv)',
# 'colors' : {'data' : r.kBlack, 'mc' : mcColor(lepton)},
# 'markers': {'data' : r.kFullCircle, 'mc' : mcMarker(lepton)},
# 'labels' : {'data' : 'Data: Conversion CR',
# 'mc' : 'MC Comb: Conv CR'}}
# plotHistRatioAndFit({'data':eff_da, 'mc':eff_mc}, ratio, fitFunc, outdir+lepton+'_fakeconv',
# graphics)
return p0
def computeAndPlotConvSf(fileData, fileMc, lepton, variable_name, outdir, outfile=None) :
"Electron conversion: simplest case, just data/mc"
eff_da = buildRate(fileData, lepton+'_fakeConv_'+variable_name)
eff_mc = buildRate(fileMc, lepton+'_fakeConv_'+variable_name)
ratio = buildRatioHistogram(eff_da, eff_mc)
print ratio.GetName(),' : ',["%.3f"%ratio.GetBinContent(b) for b in range(1, 1+ratio.GetNbinsX())]
fitFunc = fitWithConst(ratio)
p0, p0Err, chi2, ndf = fitResults(fitFunc)
p0, p0Err = pdgRound(p0, p0Err)
print "SF for %s fake conv : %s +/- %s"%(lepton, p0, p0Err)
graphics = {'xtitle' : xTitle(lepton, variable_name),
'ytitle' : lepton+' p(tight | fake conv)',
'colors' : {'data' : r.kBlack, 'mc' : mcColor(lepton)},
'markers': {'data' : r.kFullCircle, 'mc' : mcMarker(lepton)},
'labels' : {'data' : 'Data: Conversion CR',
'mc' : 'MC Comb: Conv CR'}}
plotHistRatioAndFit({'data':eff_da, 'mc':eff_mc}, ratio, fitFunc, outdir+'/fit_'+lepton+'_conv', graphics)
if outfile : saveObject(outfile, ratio, 'elec_convSF_pt')
return p0
def buildRatioHistos(histosNum={}, histosDen={}) :
"assume that the histos are organized in two dict[vt][sample] provided by buildHistos"
histosPerVtypePerSample = {}
def sameLists(l1=[], l2=[]) : return len(l1)==len(l2) and sorted(l1)==sorted(l2)
assert sameLists(histosNum.keys(), histosDen.keys()),"num and den w/ different vtype keys"
result = {}
for vt in histosDen.keys() :
resultPerSample = {}
hvn, hvd = histosNum[vt], histosDen[vt]
assert sameLists(hvn.keys(), hvd.keys()),"num and den w/ different sample keys"
for s in hvd.keys() :
hnum, hden = hvn[s], hvd[s]
hNames = [h.GetName() for h in hnum, hden]
pre, suf = commonPrefix(hNames), commonSuffix(hNames)
hName = pre+'_ratio_'+suf if len(pre) and len(suf) else '_over_'.join(hNames)
h = buildRatioHistogram(hnum, hden, hName)
resetErrors(h)
resultPerSample[s] = h
result[vt] = resultPerSample
return result
def buildRatioHistos(histosNum={}, histosDen={}) :
"assume that the histos are organized in two dict[vt][sample] provided by buildHistos"
histosPerVtypePerSample = {}
def sameLists(l1=[], l2=[]) : return len(l1)==len(l2) and sorted(l1)==sorted(l2)
assert sameLists(histosNum.keys(), histosDen.keys()),"num and den w/ different vtype keys"
result = {}
for vt in histosDen.keys() :
resultPerSample = {}
hvn, hvd = histosNum[vt], histosDen[vt]
assert sameLists(hvn.keys(), hvd.keys()),"num and den w/ different sample keys"
for s in hvd.keys() :
hnum, hden = hvn[s], hvd[s]
hNames = [h.GetName() for h in hnum, hden]
pre, suf = commonPrefix(hNames), commonSuffix(hNames)
hName = pre+'_ratio_'+suf if len(pre) and len(suf) else '_over_'.join(hNames)
h = buildRatioHistogram(hnum, hden, hName)
resetErrors(h)
resultPerSample[s] = h
result[vt] = resultPerSample
return result
def buildSideBandSubRate(file, lepton, variable_name) :
hZwindow = getNumDenHistos(file, lepton+'_realCR_' +variable_name)
hSideLo = getNumDenHistos(file, lepton+'_realSideLow_' +variable_name)
hSideHi = getNumDenHistos(file, lepton+'_realSideHigh_'+variable_name)
def sbErr(el, ew, eh) :
"SidebandError; DG this doesn't make any sense to me; ask Matt"
return sqrt(ew*ew - (el*el + eh*eh))
def be(h) : return [h.GetBinError(b) for b in range(1, 1+h.GetNbinsX())]
errs = dict([(k, [sbErr(l,w,h)
for l, w, h in zip(be(hSideLo[k]), be(hZwindow[k]), be(hSideHi[k]))])
for k in ['num','den']])
hSideLo ['num'].Add(hSideHi['num'])
hSideLo ['den'].Add(hSideHi['den'])
hZwindow['num'].Add(hSideLo['num'], -1)
hZwindow['den'].Add(hSideLo['den'], -1)
for nd in ['num', 'den'] :
h, err = hZwindow[nd], errs[nd]
nbins = h.GetNbinsX()
for i, b in zip(range(nbins), range(1, 1+nbins)) :
h.SetBinError(b, err[i])
return buildRatioHistogram(hZwindow['num'], hZwindow['den'])
def buildRatio(inputFile=None, histoBaseName='') :
num, den = inputFile.Get(histoBaseName+'_num'), inputFile.Get(histoBaseName+'_den')
return buildRatioHistogram(num, den, histoBaseName +'_rat')
def plotHistosRatio(histosPairs=[], canvasName='') :
histosRatio = [buildRatioHistogram(hh['num'], hh['den']) for hh in histosPairs]
plotHistos(histosRatio, canvasName)
def main() :
parser = optparse.OptionParser(usage=usage)
parser.add_option('-n', '--n_iter', type='int', default=8)
parser.add_option('-m', '--input_mc')
parser.add_option('-d', '--input_data')
parser.add_option('-o', '--output')
parser.add_option('-p', '--plot', help='plot inputs') # todo: implement sanity plot vs. n_iter
parser.add_option('-v','--verbose', action='store_true', default=False)
(opts, args) = parser.parse_args()
requiredOptions = ['n_iter', 'input_mc', 'input_data', 'output']
otherOptions = ['plot', 'verbose']
allOptions = requiredOptions + otherOptions
def optIsNotSpecified(o) : return not hasattr(opts, o) or getattr(opts,o) is None
if any(optIsNotSpecified(o) for o in requiredOptions) : parser.error('Missing required option')
nIter = opts.n_iter
fnameInputMc = opts.input_mc
fnameInputDa = opts.input_data
fnameOutput = opts.output
plot = opts.plot
verbose = opts.verbose
if verbose : print ('\nUsing the following options:\n'
+'\n'.join("%s : %s"%(o, str(getattr(opts, o))) for o in allOptions))
fileData = r.TFile.Open(fnameInputDa)
fileMc = r.TFile.Open(fnameInputMc)
assert fileData and fileMc, "Missing input files: data %s, mc %s"%(str(fileData), str(fileMc))
correctionHistos = {}
for lep in ['muon', 'elec'] :
if verbose : print "Lepton: %s"%lep
hRealDataCr = getNumDenHistos(fileData, lep+'_realCR_all_l_pt')
hFakeDataLo = getNumDenHistos(fileData, lep+'_fakeHF_all_l_pt')
hFakeDataHi = getNumDenHistos(fileData, lep+'_fakeHF_high_all_l_pt')
hFakeMcLo = getNumDenHistos(fileMc, lep+'_fakeHF_all_l_pt')
hFakeMcHi = getNumDenHistos(fileMc, lep+'_fakeHF_high_all_l_pt')
if plot :
hNumDen = [hFakeDataLo, hFakeDataHi, hFakeMcLo, hFakeMcHi]
for nd in ['num','den'] : plotHistos([h[nd] for h in hNumDen], 'c_'+lep+'_'+nd)
plotHistosRatio(hNumDen, 'c_'+lep+'_ratio')
def missingInputHisto(ndHistos) : return any(not h for h in ndHistos.values())
histoCollToBeChecked = ['hRealDataCr','hFakeDataLo','hFakeDataHi','hFakeMcLo','hFakeMcHi']
missingHistos = dict([(nhc,hp) for nhc,hp in [(hc, eval(hc)) for hc in histoCollToBeChecked]
if missingInputHisto(hp)])
if len(missingHistos) :
print (lep+' : missing histograms: \n'
+'\n'.join(["%s: num %s den %s"%(k, v['num'], v['den'])
for k,v in missingHistos.iteritems()]))
continue
hRealEff = buildRatioHistogram(hRealDataCr['num'], hRealDataCr['den'], 'real_eff')
corrected = dict([(nd, hFakeDataLo[nd].Clone('corrected_'+nd)) for nd in ['num', 'den']])
for iteration in range(nIter) :
rate = buildRatioHistogram(corrected['num'], corrected['den']) # temporary rate (?)
if verbose :
def lf2s(l) : return ', '.join(["%.3f"%e for e in l])
print "Iteration %d, corrected values:"%iteration
print " num %s"%lf2s(binContents(corrected['num']))
print " den %s"%lf2s(binContents(corrected['den']))
print " ratio %s"%lf2s(binContents(rate))
dataNum, dataDen = hFakeDataHi['num'], hFakeDataHi['den']
for nd,tl in [('num','tight'), ('den','loose')] :
corr, dataLow = corrected[nd], hFakeDataLo[nd]
mcLow, mcHi = hFakeMcLo[nd], hFakeMcHi[nd]
corrFact = getCorrFactors(hRealEff, rate, dataNum, dataDen, mcHi, tl)
corr = correctRate(corr, dataLow, mcLow, corrFact)
ratio = buildRatioHistogram(corrected['num'], corrected['den'], lep+'_corHFRate')
correctionHistos[lep] = ratio
if verbose : print "saving output to ",fnameOutput
fileOut = r.TFile.Open(fnameOutput, 'recreate')
fileOut.cd()
for l,h in correctionHistos.iteritems() :
if verbose : print "%s : writing %s\n%s"%(l, h.GetName(),histo1dToTxt(h))
h.Write()
fileOut.Close()
def main() :
parser = argparse.ArgumentParser(description=description,
epilog=epilog,
formatter_class=argparse.RawDescriptionHelpFormatter)
add_arg = parser.add_argument
add_arg('-o', '--output-dir', default="./")
add_arg('-p', '--process', help='one physics process, eg. ttw')
add_arg('-s', '--systematic', help='one of the systematic variations')
add_arg('-v', '--verbose', action='store_true')
add_arg('-d', '--debug', action='store_true')
args = parser.parse_args()
set_log(args.verbose, args.debug)
outdir = args.output_dir
process = args.process
systematic = args.systematic
available_processes = get_input_samples().keys()
if process not in available_processes:
raise StandardError("invalid process %s, should be one of %s"%(process, str(available_processes)))
available_systematics = get_input_samples()[process].keys()
if systematic not in available_systematics:
raise StandardError("invalid systematic %s, should be one of %s"%(systematic, str(available_systematics)))
file_label = process+'_sys_'+systematic
plot_label = process+' sys. '+systematic
normalize_to_unity = False # True
luminosity = 1.0
combiner = HistogramCombiner()
combiner.build_samples(process=process, systematic=systematic)
histogram_names = ['h_meff', 'h_jetN',
# 'h_electronPt', 'h_muonPt',
# 'h_bjetEmulN', 'h_bjetN',
# 'h_bjetEmulN_sr3b', 'h_bjetN_sr3b',
# 'h_bjetEmulN_sr1b', 'h_bjetN_sr1b',
# 'h_bjetEmulN_cr2bttV', 'h_bjetN_cr2bttV',
# 'h_meff_sr3b', 'h_jetN_sr3b', 'h_met_sr3b',
# 'h_meff_sr1b', 'h_jetN_sr1b', 'h_met_sr1b',
# 'h_meff_sr0b5j', 'h_jetN_sr0b5j', 'h_met_sr0b5j',
# 'h_meff_sr0b4j', 'h_jetN_sr0b4j', 'h_met_sr0b4j',
# 'h_meff_sr0b3j', 'h_jetN_sr0b3j', 'h_met_sr0b3j',
# 'h_meff_cr2bttV', 'h_jetN_cr2bttV', 'h_met_cr2bttV'
]
combiner.compute_normalization_factors()
output_pdf_name = outdir+'/'+file_label+'.pdf'
c_summary = R.TCanvas('c_summary', 'plotExplicitSamples sampes summary ')
combiner.print_sample_summary_to_pdf(c_summary, label="%s: nominal vs. %s systematic"%(process, systematic))
c_summary.SaveAs(output_pdf_name+'(')
for histogram_name in histogram_names:
rebin = 'meff' in histogram_name # and '_sr' in histogram_name # non-inclusive histos: low stats
rebin_factor = (2 if 'meff' in histogram_name else 2 if 'jetN' in histogram_name else 1) if rebin else 1
histograms = combiner.get_histograms(histogram_name=histogram_name)
h_nom = histograms['nominal']
h_up = histograms['up']
h_dn = histograms['down']
if 'h_jetFlavorMultiplicity' in histogram_name:
h_nom = emulate_btag_multiplicity_from_truth_flavor(h_nom, 'nom')
h_up = emulate_btag_multiplicity_from_truth_flavor(h_up, 'up')
h_dn = emulate_btag_multiplicity_from_truth_flavor(h_dn, 'dn')
histos = [h_nom, h_up, h_dn]
for h in set(histos): # set: avoid rebinning twice when up==down
h.Rebin(rebin_factor)
h_nom.SetLineWidth(2*h_nom.GetLineWidth())
h_nom.SetLineColor(R.kBlack)
h_up.SetLineColor(R.kBlue)
h_dn.SetLineColor(R.kRed)
pad_master = h_nom
pad_master.SetMaximum(1.1*max([h.GetMaximum() for h in histos]))
pad_master.SetMinimum(1.0*min([0.0]+[h.GetMinimum() for h in histos]))
pad_master.GetYaxis().SetTitle('Arbitrary Units')
pad_master.SetStats(0)
can = R.TCanvas('c_ttV_syst_'+histogram_name, 'ttV explicit variations '+pad_master.GetTitle(), 700, 700)
botPad, topPad = ru.buildBotTopPads(can, squeezeMargins=False)
# top
can.cd()
topPad.Draw()
topPad.cd()
topPad._po = [pad_master] # persistent objects
pad_master.GetXaxis().SetTitleSize(0)
pad_master.GetXaxis().SetLabelSize(0)
pad_master.Draw('axis')
# ru.topRightLabel(topPad, pad_master.GetTitle(), xpos=0.5)
ru.topRightLabel(topPad, "#bf{#it{ATLAS}} Simulation Preliminary", xpos=0.85, ypos=0.9)
ru.topRightLabel(topPad, "#sqrt{s} = 13 TeV", xpos=0.85, ypos=0.8)
leg = ru.topRightLegend(can, legWidth=0.225, legHeight=0.300, hShift=-0.10, vShift=-0.25)
leg.SetBorderSize(0)
# leg.SetHeader(plot_label+ ("(norm=1)" if normalize_to_unity else "(lumi %.1f)"%luminosity))
topPad._po.append(leg)
def format_legend_label(h, l):
return "{0}: {1:.2E} ({2:.0f})".format(l, h.Integral(), h.GetEntries())
def pretty_scale_legend_label(h, l):
return ("nominal" if l is 'nom' else
"#mu = 2.0 #mu_{0}" if l is 'up' else
"#mu = 0.5 #mu_{0}" if l is 'dn' else
'unknown')
for h,l in [(h_nom, 'nom'), (h_up, 'up'), (h_dn, 'dn')]:
h.Draw('hist same')
leg.AddEntry(h, pretty_scale_legend_label(h, l), 'l')
topPad._po.append(h)
leg.Draw('same')
def integral_and_error(h):
error = R.Double(0.0)
integral = h.IntegralAndError(1, h.GetNbinsX()+1, error)
return integral, error
def ratio_and_error(ave=(1.0, 0.01), bve=(2.0, 0.001)):
a, sa = ave
b, sb = bve
if a and b:
r = a/b
e = r * sqrt((sa/a)*(sa/a)+(sb/b)*(sb/b))
return r, e
else:
return 0.0, 0.0
print_normalization_summary = histogram_name.startswith('h_meff')
if print_normalization_summary:
nom_int = h_nom.Integral()
up_int = h_up.Integral()
dn_int = h_dn.Integral()
nom_int, nom_err = integral_and_error(h_nom)
up_int, up_err = integral_and_error(h_up)
dn_int, dn_err = integral_and_error(h_dn)
rup, rupe = ratio_and_error((up_int, up_err), (nom_int, nom_err))
rdn, rdne = ratio_and_error((dn_int, dn_err), (nom_int, nom_err))
# print ("normalization change: "
# +"{} up {:.1%} down {:.1%} (nom {:.1f}, up {:.1f}, do {:.1f})".format(h_nom.GetName(),
# 1.0-up_int/nom_int if nom_int else 1.0,
# 1.0-dn_int/nom_int if nom_int else 1.0,
# nom_int,
# up_int,
# dn_int))
print ("normalization change: "
+"{} up {:.1%} +/- {:.1%} down {:.1%} +/- {:.1%} ".format(h_nom.GetName(), 1.0-rup, rupe, 1.0-rdn, rdne)
+"(integral: "
+"nom {:.2E} +/- {:.2E}, up {:.2E} +/- {:.2E}, do {:.2E} +/- {:.2E})".format(nom_int, nom_err,
up_int, up_err,
dn_int, dn_err)
+" (entries nom {:.2E} up {:.2E} do {:.2E}".format(h_nom.GetEntries(), h_up.GetEntries(), h_dn.GetEntries()))
print ("tex normalization change: "
+"{} up ${:.1%} \pm {:.1%}$ down ${:.1%} \pm {:.1%}$ ".format(h_nom.GetName(), 1.0-rup, rupe, 1.0-rdn, rdne)
)
def bc(h): return [h.GetBinContent(i) for i in range(1,1+h.GetNbinsX())]
def max_frac_variation(h1, h2):
"maximum bin-by-bin fractional variation; h1 is denominator, empty bins skipped"
bc1 = bc(h1)
bc2 = bc(h2)
return max([abs(b2/b1) for b1, b2 in zip(bc1, bc2) if b1 and b2])
def max_frac_variation_within10(h1, h2):
"""maximum bin-by-bin fractional variation; h1 is denominator.
Bins with <0.1*peak are skipped"""
bc1 = bc(h1)
bc2 = bc(h2)
m1 = max(bc1)
m2 = max(bc2)
return max([abs(b2/b1) for b1, b2 in zip(bc1, bc2) if b1>0.1*m1 and b2>0.1*m2])
# print ("shape change: "
# +"{} up {:.1%} down {:.1%} ".format(h_nom.GetName(),
# 1.0-max_frac_variation_within10(h_up, h_nom),
# 1.0-max_frac_variation_within10(h_dn, h_nom)))
topPad.Update()
# bottom
can.cd()
botPad.SetTopMargin(1.25*botPad.GetTopMargin())
botPad.Draw()
botPad.cd()
ratio_up = ru.buildRatioHistogram(h_up, h_nom)
ratio_dn = ru.buildRatioHistogram(h_dn, h_nom)
yMin, yMax = 0.5, 1.5
ratioPadMaster = pad_master.Clone(pad_master.GetName()+'_ratio')
ratioPadMaster.SetMinimum(yMin)
ratioPadMaster.SetMaximum(yMax)
ratioPadMaster.SetStats(0)
ratioPadMaster.Draw('axis')
x_lo, x_hi = ru.getXrange(ratioPadMaster)
refLines = [ru.referenceLine(x_lo, x_hi, y, y) for y in [0.5, 1.0, 1.5] if y>yMin and y<yMax]
for l in refLines : l.Draw()
ratio_up.Draw('same')
ratio_dn.Draw('same')
xA, yA = ratioPadMaster.GetXaxis(), ratioPadMaster.GetYaxis()
textScaleUp = 0.75*1.0/botPad.GetHNDC()
yA.SetNdivisions(-102)
yA.SetTitle('Ratio')
yA.CenterTitle()
yA.SetTitleOffset(yA.GetTitleOffset()/textScaleUp)
xA.SetTitleSize(yA.GetTitleSize()) # x- was set to 0 for padmaster, restore it
xA.SetLabelSize(yA.GetLabelSize())
xA.SetTitle(prettify_title(xA.GetTitle()))
for a in [xA, yA] :
a.SetLabelSize(a.GetLabelSize()*textScaleUp)
a.SetTitleSize(a.GetTitleSize()*textScaleUp)
botPad._graphical_objects = [ratio_up, ratio_dn, ratioPadMaster] + refLines # avoid garbage collection
botPad.Update()
can.Update()
first_histo = histogram_name is histogram_names[0]
last_histo = histogram_name is histogram_names[-1]
can.SaveAs(outdir+'/'+can.GetName()+'.png')
can.SaveAs(outdir+'/'+can.GetName()+'.eps')
can.SaveAs(output_pdf_name+ (')' if last_histo else ''))
def buildRate(file, histo_basename) :
hs = getNumDenHistos(file, histo_basename)
return buildRatioHistogram(hs['num'], hs['den'])
def buildRatio(inputFile=None, histoBaseName=''):
num, den = inputFile.Get(histoBaseName +
'_num'), inputFile.Get(histoBaseName + '_den')
return buildRatioHistogram(num, den, histoBaseName + '_rat')
def buildRate(file, histo_basename):
hs = getNumDenHistos(file, histo_basename)
return buildRatioHistogram(hs['num'], hs['den'])
def plotHistosRatio(histosPairs=[], canvasName=''):
histosRatio = [
buildRatioHistogram(hh['num'], hh['den']) for hh in histosPairs
]
plotHistos(histosRatio, canvasName)
def main():
parser = argparse.ArgumentParser(
description=description,
epilog=epilog,
formatter_class=argparse.RawDescriptionHelpFormatter)
add_arg = parser.add_argument
add_arg('-o', '--output-dir', default="./")
add_arg('-p', '--process', help='one physics process, eg. ttw')
add_arg('-s', '--systematic', help='one of the systematic variations')
add_arg('-v', '--verbose', action='store_true')
add_arg('-d', '--debug', action='store_true')
args = parser.parse_args()
set_log(args.verbose, args.debug)
outdir = args.output_dir
process = args.process
systematic = args.systematic
available_processes = get_input_samples().keys()
if process not in available_processes:
raise StandardError("invalid process %s, should be one of %s" %
(process, str(available_processes)))
available_systematics = get_input_samples()[process].keys()
if systematic not in available_systematics:
raise StandardError("invalid systematic %s, should be one of %s" %
(systematic, str(available_systematics)))
file_label = process + '_sys_' + systematic
plot_label = process + ' sys. ' + systematic
normalize_to_unity = False # True
luminosity = 1.0
combiner = HistogramCombiner()
combiner.build_samples(process=process, systematic=systematic)
histogram_names = [
'h_meff',
'h_jetN',
# 'h_electronPt', 'h_muonPt',
# 'h_bjetEmulN', 'h_bjetN',
# 'h_bjetEmulN_sr3b', 'h_bjetN_sr3b',
# 'h_bjetEmulN_sr1b', 'h_bjetN_sr1b',
# 'h_bjetEmulN_cr2bttV', 'h_bjetN_cr2bttV',
# 'h_meff_sr3b', 'h_jetN_sr3b', 'h_met_sr3b',
# 'h_meff_sr1b', 'h_jetN_sr1b', 'h_met_sr1b',
# 'h_meff_sr0b5j', 'h_jetN_sr0b5j', 'h_met_sr0b5j',
# 'h_meff_sr0b4j', 'h_jetN_sr0b4j', 'h_met_sr0b4j',
# 'h_meff_sr0b3j', 'h_jetN_sr0b3j', 'h_met_sr0b3j',
# 'h_meff_cr2bttV', 'h_jetN_cr2bttV', 'h_met_cr2bttV'
]
combiner.compute_normalization_factors()
output_pdf_name = outdir + '/' + file_label + '.pdf'
c_summary = R.TCanvas('c_summary', 'plotExplicitSamples sampes summary ')
combiner.print_sample_summary_to_pdf(
c_summary,
label="%s: nominal vs. %s systematic" % (process, systematic))
c_summary.SaveAs(output_pdf_name + '(')
for histogram_name in histogram_names:
rebin = 'meff' in histogram_name # and '_sr' in histogram_name # non-inclusive histos: low stats
rebin_factor = (2 if 'meff' in histogram_name else
2 if 'jetN' in histogram_name else 1) if rebin else 1
histograms = combiner.get_histograms(histogram_name=histogram_name)
h_nom = histograms['nominal']
h_up = histograms['up']
h_dn = histograms['down']
if 'h_jetFlavorMultiplicity' in histogram_name:
h_nom = emulate_btag_multiplicity_from_truth_flavor(h_nom, 'nom')
h_up = emulate_btag_multiplicity_from_truth_flavor(h_up, 'up')
h_dn = emulate_btag_multiplicity_from_truth_flavor(h_dn, 'dn')
histos = [h_nom, h_up, h_dn]
for h in set(histos): # set: avoid rebinning twice when up==down
h.Rebin(rebin_factor)
h_nom.SetLineWidth(2 * h_nom.GetLineWidth())
h_nom.SetLineColor(R.kBlack)
h_up.SetLineColor(R.kBlue)
h_dn.SetLineColor(R.kRed)
pad_master = h_nom
pad_master.SetMaximum(1.1 * max([h.GetMaximum() for h in histos]))
pad_master.SetMinimum(1.0 *
min([0.0] + [h.GetMinimum() for h in histos]))
pad_master.GetYaxis().SetTitle('Arbitrary Units')
pad_master.SetStats(0)
can = R.TCanvas('c_ttV_syst_' + histogram_name,
'ttV explicit variations ' + pad_master.GetTitle(),
700, 700)
botPad, topPad = ru.buildBotTopPads(can, squeezeMargins=False)
# top
can.cd()
topPad.Draw()
topPad.cd()
topPad._po = [pad_master] # persistent objects
pad_master.GetXaxis().SetTitleSize(0)
pad_master.GetXaxis().SetLabelSize(0)
pad_master.Draw('axis')
# ru.topRightLabel(topPad, pad_master.GetTitle(), xpos=0.5)
ru.topRightLabel(topPad,
"#bf{#it{ATLAS}} Simulation Preliminary",
xpos=0.85,
ypos=0.9)
ru.topRightLabel(topPad, "#sqrt{s} = 13 TeV", xpos=0.85, ypos=0.8)
leg = ru.topRightLegend(can,
legWidth=0.225,
legHeight=0.300,
hShift=-0.10,
vShift=-0.25)
leg.SetBorderSize(0)
# leg.SetHeader(plot_label+ ("(norm=1)" if normalize_to_unity else "(lumi %.1f)"%luminosity))
topPad._po.append(leg)
def format_legend_label(h, l):
return "{0}: {1:.2E} ({2:.0f})".format(l, h.Integral(),
h.GetEntries())
def pretty_scale_legend_label(h, l):
return ("nominal"
if l is 'nom' else "#mu = 2.0 #mu_{0}" if l is 'up' else
"#mu = 0.5 #mu_{0}" if l is 'dn' else 'unknown')
for h, l in [(h_nom, 'nom'), (h_up, 'up'), (h_dn, 'dn')]:
h.Draw('hist same')
leg.AddEntry(h, pretty_scale_legend_label(h, l), 'l')
topPad._po.append(h)
leg.Draw('same')
def integral_and_error(h):
error = R.Double(0.0)
integral = h.IntegralAndError(1, h.GetNbinsX() + 1, error)
return integral, error
def ratio_and_error(ave=(1.0, 0.01), bve=(2.0, 0.001)):
a, sa = ave
b, sb = bve
if a and b:
r = a / b
e = r * sqrt((sa / a) * (sa / a) + (sb / b) * (sb / b))
return r, e
else:
return 0.0, 0.0
print_normalization_summary = histogram_name.startswith('h_meff')
if print_normalization_summary:
nom_int = h_nom.Integral()
up_int = h_up.Integral()
dn_int = h_dn.Integral()
nom_int, nom_err = integral_and_error(h_nom)
up_int, up_err = integral_and_error(h_up)
dn_int, dn_err = integral_and_error(h_dn)
rup, rupe = ratio_and_error((up_int, up_err), (nom_int, nom_err))
rdn, rdne = ratio_and_error((dn_int, dn_err), (nom_int, nom_err))
# print ("normalization change: "
# +"{} up {:.1%} down {:.1%} (nom {:.1f}, up {:.1f}, do {:.1f})".format(h_nom.GetName(),
# 1.0-up_int/nom_int if nom_int else 1.0,
# 1.0-dn_int/nom_int if nom_int else 1.0,
# nom_int,
# up_int,
# dn_int))
print(
"normalization change: " +
"{} up {:.1%} +/- {:.1%} down {:.1%} +/- {:.1%} ".format(
h_nom.GetName(), 1.0 - rup, rupe, 1.0 - rdn, rdne) +
"(integral: " +
"nom {:.2E} +/- {:.2E}, up {:.2E} +/- {:.2E}, do {:.2E} +/- {:.2E})"
.format(nom_int, nom_err, up_int, up_err, dn_int, dn_err) +
" (entries nom {:.2E} up {:.2E} do {:.2E}".format(
h_nom.GetEntries(), h_up.GetEntries(), h_dn.GetEntries()))
print(
"tex normalization change: " +
"{} up ${:.1%} \pm {:.1%}$ down ${:.1%} \pm {:.1%}$ ".format(
h_nom.GetName(), 1.0 - rup, rupe, 1.0 - rdn, rdne))
def bc(h):
return [
h.GetBinContent(i) for i in range(1, 1 + h.GetNbinsX())
]
def max_frac_variation(h1, h2):
"maximum bin-by-bin fractional variation; h1 is denominator, empty bins skipped"
bc1 = bc(h1)
bc2 = bc(h2)
return max(
[abs(b2 / b1) for b1, b2 in zip(bc1, bc2) if b1 and b2])
def max_frac_variation_within10(h1, h2):
"""maximum bin-by-bin fractional variation; h1 is denominator.
Bins with <0.1*peak are skipped"""
bc1 = bc(h1)
bc2 = bc(h2)
m1 = max(bc1)
m2 = max(bc2)
return max([
abs(b2 / b1) for b1, b2 in zip(bc1, bc2)
if b1 > 0.1 * m1 and b2 > 0.1 * m2
])
# print ("shape change: "
# +"{} up {:.1%} down {:.1%} ".format(h_nom.GetName(),
# 1.0-max_frac_variation_within10(h_up, h_nom),
# 1.0-max_frac_variation_within10(h_dn, h_nom)))
topPad.Update()
# bottom
can.cd()
botPad.SetTopMargin(1.25 * botPad.GetTopMargin())
botPad.Draw()
botPad.cd()
ratio_up = ru.buildRatioHistogram(h_up, h_nom)
ratio_dn = ru.buildRatioHistogram(h_dn, h_nom)
yMin, yMax = 0.5, 1.5
ratioPadMaster = pad_master.Clone(pad_master.GetName() + '_ratio')
ratioPadMaster.SetMinimum(yMin)
ratioPadMaster.SetMaximum(yMax)
ratioPadMaster.SetStats(0)
ratioPadMaster.Draw('axis')
x_lo, x_hi = ru.getXrange(ratioPadMaster)
refLines = [
ru.referenceLine(x_lo, x_hi, y, y) for y in [0.5, 1.0, 1.5]
if y > yMin and y < yMax
]
for l in refLines:
l.Draw()
ratio_up.Draw('same')
ratio_dn.Draw('same')
xA, yA = ratioPadMaster.GetXaxis(), ratioPadMaster.GetYaxis()
textScaleUp = 0.75 * 1.0 / botPad.GetHNDC()
yA.SetNdivisions(-102)
yA.SetTitle('Ratio')
yA.CenterTitle()
yA.SetTitleOffset(yA.GetTitleOffset() / textScaleUp)
xA.SetTitleSize(
yA.GetTitleSize()) # x- was set to 0 for padmaster, restore it
xA.SetLabelSize(yA.GetLabelSize())
xA.SetTitle(prettify_title(xA.GetTitle()))
for a in [xA, yA]:
a.SetLabelSize(a.GetLabelSize() * textScaleUp)
a.SetTitleSize(a.GetTitleSize() * textScaleUp)
botPad._graphical_objects = [ratio_up, ratio_dn, ratioPadMaster
] + refLines # avoid garbage collection
botPad.Update()
can.Update()
first_histo = histogram_name is histogram_names[0]
last_histo = histogram_name is histogram_names[-1]
can.SaveAs(outdir + '/' + can.GetName() + '.png')
can.SaveAs(outdir + '/' + can.GetName() + '.eps')
can.SaveAs(output_pdf_name + (')' if last_histo else ''))
def plotHistosRatio(histosPairs=[], canvasName="", outdir="./"):
histosRatio = [buildRatioHistogram(hh["num"], hh["den"]) for hh in histosPairs]
print histosRatio
for h in histosRatio:
h.SetTitle("ratio " + h.GetTitle())
plotHistos(histosRatio, canvasName, outdir)
def main():
parser = optparse.OptionParser(usage=usage)
parser.add_option('-n', '--n_iter', type='int', default=8)
parser.add_option('-m', '--input_mc')
parser.add_option('-d', '--input_data')
parser.add_option('-o', '--output')
parser.add_option(
'-p', '--plot',
help='plot inputs') # todo: implement sanity plot vs. n_iter
parser.add_option('-v', '--verbose', action='store_true', default=False)
(opts, args) = parser.parse_args()
requiredOptions = ['n_iter', 'input_mc', 'input_data', 'output']
otherOptions = ['plot', 'verbose']
allOptions = requiredOptions + otherOptions
def optIsNotSpecified(o):
return not hasattr(opts, o) or getattr(opts, o) is None
if any(optIsNotSpecified(o) for o in requiredOptions):
parser.error('Missing required option')
nIter = opts.n_iter
fnameInputMc = opts.input_mc
fnameInputDa = opts.input_data
fnameOutput = opts.output
plot = opts.plot
verbose = opts.verbose
if verbose:
print(
'\nUsing the following options:\n' +
'\n'.join("%s : %s" % (o, str(getattr(opts, o)))
for o in allOptions))
fileData = r.TFile.Open(fnameInputDa)
fileMc = r.TFile.Open(fnameInputMc)
assert fileData and fileMc, "Missing input files: data %s, mc %s" % (
str(fileData), str(fileMc))
correctionHistos = {}
for lep in ['muon', 'elec']:
if verbose: print "Lepton: %s" % lep
hRealDataCr = getNumDenHistos(fileData, lep + '_realCR_all_l_pt')
hFakeDataLo = getNumDenHistos(fileData, lep + '_fakeHF_all_l_pt')
hFakeDataHi = getNumDenHistos(fileData, lep + '_fakeHF_high_all_l_pt')
hFakeMcLo = getNumDenHistos(fileMc, lep + '_fakeHF_all_l_pt')
hFakeMcHi = getNumDenHistos(fileMc, lep + '_fakeHF_high_all_l_pt')
if plot:
hNumDen = [hFakeDataLo, hFakeDataHi, hFakeMcLo, hFakeMcHi]
for nd in ['num', 'den']:
plotHistos([h[nd] for h in hNumDen], 'c_' + lep + '_' + nd)
plotHistosRatio(hNumDen, 'c_' + lep + '_ratio')
def missingInputHisto(ndHistos):
return any(not h for h in ndHistos.values())
histoCollToBeChecked = [
'hRealDataCr', 'hFakeDataLo', 'hFakeDataHi', 'hFakeMcLo',
'hFakeMcHi'
]
missingHistos = dict([(nhc, hp)
for nhc, hp in [(hc, eval(hc))
for hc in histoCollToBeChecked]
if missingInputHisto(hp)])
if len(missingHistos):
print(
lep + ' : missing histograms: \n' + '\n'.join([
"%s: num %s den %s" % (k, v['num'], v['den'])
for k, v in missingHistos.iteritems()
]))
continue
hRealEff = buildRatioHistogram(hRealDataCr['num'], hRealDataCr['den'],
'real_eff')
corrected = dict([(nd, hFakeDataLo[nd].Clone('corrected_' + nd))
for nd in ['num', 'den']])
for iteration in range(nIter):
rate = buildRatioHistogram(corrected['num'],
corrected['den']) # temporary rate (?)
if verbose:
def lf2s(l):
return ', '.join(["%.3f" % e for e in l])
print "Iteration %d, corrected values:" % iteration
print " num %s" % lf2s(binContents(corrected['num']))
print " den %s" % lf2s(binContents(corrected['den']))
print " ratio %s" % lf2s(binContents(rate))
dataNum, dataDen = hFakeDataHi['num'], hFakeDataHi['den']
for nd, tl in [('num', 'tight'), ('den', 'loose')]:
corr, dataLow = corrected[nd], hFakeDataLo[nd]
mcLow, mcHi = hFakeMcLo[nd], hFakeMcHi[nd]
corrFact = getCorrFactors(hRealEff, rate, dataNum, dataDen,
mcHi, tl)
corr = correctRate(corr, dataLow, mcLow, corrFact)
ratio = buildRatioHistogram(corrected['num'], corrected['den'],
lep + '_corHFRate')
correctionHistos[lep] = ratio
if verbose: print "saving output to ", fnameOutput
fileOut = r.TFile.Open(fnameOutput, 'recreate')
fileOut.cd()
for l, h in correctionHistos.iteritems():
if verbose:
print "%s : writing %s\n%s" % (l, h.GetName(), histo1dToTxt(h))
h.Write()
fileOut.Close()
def plotIsoComparison(histosPerSource={}, outputDir='', region='', lepton='', verbose=False):
"""
plot a comparison of eff(T|L) for real and for fake leptons
vs. pt, where the numerator is one of the tight definitions
"""
var = 'pt'
sources = histosPerSource.keys()
lOrTOrTs = first(first(histosPerSource)).keys()
histosPtPerSource = dict((s, dict((lt, histosPerSource[s][var][lt]) for lt in lOrTOrTs)) for s in sources)
def buildTotFakeHistos():
"add up all the non-real (fake) sources"
notRealSources = [s for s in sources if s!='real']
aSource = first(notRealSources)
totFakeHistos = dict()
for lt in ['loose', 'tight', 'tight_std', 'tight_minden', 'tight_tight']:
template = histosPtPerSource[aSource][lt]
h = template.Clone(template.GetName().replace(aSource, 'fake'))
h.Reset()
for s in sources : h.Add(histosPtPerSource[s][lt])
totFakeHistos[lt] = h
return totFakeHistos
histosPtPerSource['fake'] = buildTotFakeHistos()
effReal_wh = rootUtils.buildRatioHistogram(histosPtPerSource['real']['tight' ], histosPtPerSource['real']['loose'])
effReal_std = rootUtils.buildRatioHistogram(histosPtPerSource['real']['tight_std' ], histosPtPerSource['real']['loose'])
effReal_minden = rootUtils.buildRatioHistogram(histosPtPerSource['real']['tight_minden'], histosPtPerSource['real']['loose'])
effReal_tight = rootUtils.buildRatioHistogram(histosPtPerSource['real']['tight_tight' ], histosPtPerSource['real']['loose'])
effFake_wh = rootUtils.buildRatioHistogram(histosPtPerSource['fake']['tight' ], histosPtPerSource['fake']['loose'])
effFake_std = rootUtils.buildRatioHistogram(histosPtPerSource['fake']['tight_std' ], histosPtPerSource['fake']['loose'])
effFake_minden = rootUtils.buildRatioHistogram(histosPtPerSource['fake']['tight_minden'], histosPtPerSource['fake']['loose'])
effFake_tight = rootUtils.buildRatioHistogram(histosPtPerSource['fake']['tight_tight' ], histosPtPerSource['fake']['loose'])
frameName, frameTitle = region+'_'+lepton, "fake and real efficiencies for %s in %s"%(lepton, region)
can = r.TCanvas('c_'+frameName, frameTitle, 800, 600)
can.cd()
pm = effReal_wh
pm.SetMinimum(0.0)
pm.SetMaximum(1.1)
pm.GetYaxis().SetTitle("#epsilon(T|L)")
colorReal, colorFake = r.kBlue, r.kRed
markerWh, markerStd, markerMinden, markerTight = r.kMultiply, r.kCircle, r.kOpenTriangleUp, r.kOpenSquare
def setAttrs(h, mark, col):
h.SetLineColor(col)
h.SetMarkerColor(col)
h.SetMarkerStyle(mark)
setAttrs(effReal_wh, markerWh, colorReal)
setAttrs(effReal_std, markerStd, colorReal)
setAttrs(effReal_minden, markerMinden, colorReal)
setAttrs(effReal_tight, markerTight, colorReal)
setAttrs(effFake_wh, markerWh, colorFake)
setAttrs(effFake_std, markerStd, colorFake)
setAttrs(effFake_minden, markerMinden, colorFake)
setAttrs(effFake_tight, markerTight, colorFake)
pm.SetStats(0)
pm.Draw('axis')
#for h in [effReal_wh, effReal_std, effReal_tight, effFake_wh, effFake_std, effFake_tight]:
for h in [effReal_wh, effReal_std, effReal_minden, effFake_wh, effFake_std, effFake_minden]:
h.Draw('same')
leg = rightLegend(can)
leg.SetBorderSize(0)
leg.AddEntry(r.TObject(), 'Real', '')
leg.AddEntry(effReal_std, 'std iso', 'lp')
#leg.AddEntry(effReal_tight, 'tight iso', 'lp')
leg.AddEntry(effReal_minden,'minden iso', 'lp')
leg.AddEntry(effReal_wh, 'wh iso', 'lp')
leg.AddEntry(r.TObject(), 'Fake', '')
leg.AddEntry(effFake_std, 'std iso', 'lp')
#leg.AddEntry(effFake_tight, 'tight iso', 'lp')
leg.AddEntry(effFake_minden,'minden iso', 'lp')
leg.AddEntry(effFake_wh, ' wh iso', 'lp')
leg.Draw()
topRightLabel(can, "#splitline{%s}{%s}"%(lepton, region), xpos=0.125, align=13)
can.RedrawAxis()
can._histos = [effReal_wh, effReal_std, effFake_wh, effFake_std]
can.Update()
mkdirIfNeeded(outputDir)
can.SaveAs(os.path.join(outputDir, frameTitle+'.png'))