from datasets import allsamples
import purity.selections as selections
import config
import utils
ROOT.gROOT.SetBatch(True)
###########################
# Setup and configuration #
###########################
snames = ['sph-16*-m']
lumi = sum(s.lumi for s in allsamples.getmany(snames))
canvas = SimpleCanvas(lumi=lumi)
rcanvas = RatioCanvas(lumi=lumi)
binning = array.array('d', map(float, selections.photonPtBinning))
pid = 'medium'
tune = 'Spring16' # 'GJetsCWIso'
extras = 'pixel-chargedpf' # -monoph'
suffix = '_Spring16' # '_GJetsCWIso'
itune = selections.Tunes.index(tune)
inputFile = ROOT.TFile.Open(basedir + '/data/impurity_' + tune + '.root')
impurityGraph = inputFile.Get("barrel-" + pid + "-" + extras + "-Met0to60")
outputFile = ROOT.TFile.Open(
basedir + '/data/hadronTFactor' + suffix + '.root', 'recreate')
puweight.Divide(mcDist)
for bin in range(1, puweight.GetNbinsX() + 1):
puweight.SetBinError(bin, 0.)
# HACK
if puweight.GetBinContent(bin) > 4.:
puweight.SetBinContent(bin, 4.)
puweight.Write()
else:
outputFile = ROOT.TFile.Open(sys.argv[1])
dataDist = outputFile.Get('data')
canvas = RatioCanvas('ratio')
canvas.legend.setPosition(0.7, 0.8, 0.9, 0.9)
canvas.legend.add('data',
'Data',
opt='LF',
color=ROOT.kRed,
fstyle=3003,
lwidth=2)
canvas.legend.add('mc', '', opt='LF', color=ROOT.kBlue, fstyle=3003, lwidth=2)
canvas.legend.apply('data', dataDist)
dataClone = dataDist.Clone('points')
dataClone.SetFillStyle(0)
dataClone.SetMarkerStyle(1)
dataClone.SetMarkerColor(ROOT.kBlack)
scanvas.addHistogram(hdata, drawOpt = 'EP')
scanvas.addHistogram(hmc, drawOpt = 'HIST')
scanvas.xtitle = '#sigma_{i#etai#eta}'
contam = hmc.Integral() / hdata.Integral() * 100.
text = 'Sig. Contam: {contam}%'.format(contam = round(contam,1))
scanvas.addText(text, 0.225, 0.4, 0.4, 0.6)
scanvas.printWeb(pdir + '/sbcontam' , 'sbcontam_' + name, logy = False)
plotSigContam(hDataBkgNom, hMCSBNom, name = 'nominal')
plotSigContam(hDataBkgNear, hMCSBNear, name = 'near')
plotSigContam(hDataBkgFar, hMCSBFar, name = 'far')
canvas = RatioCanvas(lumi = s.sphLumi)
# allowing the bin edge to be lower than the actual cut (makes the purity higher!)
cutBin = hDataTarg.FindBin(var.cuts[pid])
FitResult = collections.namedtuple('FitResult', ['purity', 'aveSig', 'nReal', 'nFake'])
def plotSSFit(fitter, purity, nReal, name = '', pdir = plotDir):
fitter.preparePlot()
target = fitter.getTarget()
total = fitter.getTotal()
sig = fitter.getSignal()
sigcr = fitter.getSignalCR()
bkg = fitter.getBackground()
subbkg = fitter.getSubtractedBackground()
(50, -5., 5.), '1', False),
('jgammaDPhi', '#Delta#phi(#gamma, j1);#Delta#phi;pb / (#pi/30)',
(30, 0., math.pi), 'hasMatchingPhoton && jet1Pt > 0.', False)]
samples = [('znng130MP', ('znng-130', 'znng-130-o'), 'pgammaPt > 140.'),
('wnlg130MP', ('wnlg-130', 'wnlg-130-o'), 'pgammaPt > 140.'),
('zllg130MP', ('zllg-130', 'zllg-130-o'), 'pgammaPt > 140.'),
('wnlg', ('wnlg', 'wglo'), 'pgammaPt > 0.'),
('zllg', ('zllg', None), 'pgammaPt > 0.'),
('znng', ('znng', None), 'pgammaPt > 0.'),
('wnlg130', ('wnlg-130', 'wglo-130'), 'pgammaPt > 140.'),
('wlng500', ('wnlg-500', 'wglo-500'), 'pgammaPt > 510.')]
source = ROOT.TFile.Open(config.histDir + '/kfactor_trees.root')
canvas = RatioCanvas(sim=True)
canvas.legend.setPosition(0.7, 0.7, 0.9, 0.9)
canvas.legend.add('lo',
'LO',
opt='L',
color=ROOT.kBlue,
lwidth=2,
lstyle=ROOT.kSolid)
canvas.legend.add('nlo',
'NLO',
opt='L',
color=ROOT.kRed,
lwidth=2,
lstyle=ROOT.kSolid)
for name, (nlo, lo), baseline in samples:
def plotFit(mass,
targHist,
model,
dataType,
suffix,
bkgModel='nombkgModel',
hmcbkg=None,
alt='',
plotName=''):
global canvas
if canvas is None:
canvas = RatioCanvas(lumi=lumi, sim=(dataType == 'mc'))
canvas.Clear(full=True)
canvas.titlePave.SetX2NDC(0.5)
canvas.legend.setPosition(0.6, 0.7, 0.9, 0.9)
canvas.legend.add('obs',
title='Observed',
opt='LP',
color=ROOT.kBlack,
mstyle=8)
canvas.legend.add('fit',
title='Fit',
opt='L',
lcolor=ROOT.kBlue,
lwidth=2,
lstyle=ROOT.kSolid)
canvas.legend.add('bkg',
title='Bkg component',
opt='L',
lcolor=ROOT.kGreen,
lwidth=2,
lstyle=ROOT.kDashed)
if hmcbkg:
canvas.legend.add('mcbkg',
title='Bkg (MC truth)',
opt='LF',
lcolor=ROOT.kRed,
lwidth=1,
fcolor=ROOT.kRed,
fstyle=3003)
if targHist.sumEntries() > 500.:
plotBinning = plotBinningT
else:
plotBinning = plotBinningT2
frame = mass.frame()
targHist.plotOn(frame, ROOT.RooFit.Binning(*plotBinning))
model.plotOn(frame)
model.plotOn(frame, ROOT.RooFit.Components(bkgModel + '_' + suffix),
ROOT.RooFit.LineStyle(ROOT.kDashed),
ROOT.RooFit.LineColor(ROOT.kGreen))
frame.SetTitle('')
frame.SetMinimum(0.)
canvas.addHistogram(frame, clone=True, drawOpt='')
if hmcbkg:
htruth = hmcbkg.Rebin(hmcbkg.GetNbinsX() / plotBinning[0], 'truth')
canvas.legend.apply('mcbkg', htruth)
canvas.addHistogram(htruth)
canvas.rlimits = (-2., 2.)
canvas.Update(rList=[], logy=False)
frame.Print()
# adding ratio pad
fitcurve = frame.findObject(model.GetName() + '_Norm[mass]')
hresid = targHist.createHistogram('residual', mass,
ROOT.RooFit.Binning(*plotBinning))
rdata = ROOT.TGraphErrors(hresid.GetNbinsX())
for iP in range(rdata.GetN()):
x = hresid.GetXaxis().GetBinCenter(iP + 1)
nData = hresid.GetBinContent(iP + 1)
statErr = hresid.GetBinError(iP + 1)
nFit = fitcurve.interpolate(x)
if statErr > 0.:
rdata.SetPoint(iP, x, (nData - nFit) / statErr)
else:
rdata.SetPoint(iP, x, (nData - nFit))
# rdata.SetPointError(iP, 0., dmet.GetBinError(iP + 1) / norm)
rdata.SetMarkerStyle(8)
rdata.SetMarkerColor(ROOT.kBlack)
rdata.SetLineColor(ROOT.kBlack)
canvas.ratioPad.cd()
canvas.rtitle = '(data - fit) / #sigma_{data}'
rframe = ROOT.TH1F('rframe', '', 1, *plotBinning[1:])
rframe.GetYaxis().SetRangeUser(-2., 2.)
rframe.Draw()
line = ROOT.TLine(plotBinning[1], 0., plotBinning[2], 0.)
line.SetLineWidth(2)
line.SetLineColor(ROOT.kBlue)
line.Draw()
rdata.Draw('EP')
canvas._needUpdate = False
if not plotName:
plotName = 'fit_' + dataType + '_' + suffix
if alt:
plotName += '_' + alt
canvas.printWeb(plotDir, plotName, logy=False)
rframe.Delete()
if hmcbkg:
htruth.Delete()
hresid.Delete()
for iX in range(hnominal.GetNbinsX() + 2):
if ROOT.TMath.IsNaN(hnominal.GetBinContent(iX)) or ROOT.TMath.IsNaN(hnominal.GetBinError(iX)):
print hnominal.GetName(), 'cannot be plotted because it contains NaN'
plot = False
break
if huncert and (ROOT.TMath.IsNaN(huncert.GetBinContent(iX)) or ROOT.TMath.IsNaN(huncert.GetBinError(iX))):
print huncert.GetName(), 'cannot be plotted because it contains NaN'
plot = False
break
if not plot:
continue
canvas1.legend.apply('stat', hnominal)
if huncert:
canvas1.Clear()
canvas1.legend.apply('total', huncert)
canvas1.addHistogram(huncert, drawOpt = 'E2')
canvas1.addHistogram(hnominal, drawOpt = 'EP')
canvas1.printWeb(plotDir, hnominal.GetName(), logy = not proc.startswith('tf_'))
else:
canvas2.Clear()
canvas2.addHistogram(hnominal, drawOpt = 'EP')
canvas2.printWeb(plotDir, hnominal.GetName(), logy = not proc.startswith('tf_'))
rcanvas = RatioCanvas(name = 'datamc', lumi = 36400)
rcanvas.legend.setPosition(0.7, 0.7, 0.9, 0.9)
floatName = matches.group(1)
if len(floatNames) != 0:
postfitSub = postfitTotal.Clone('subdominant')
for proc in floatNames:
postfitSub.Add(postfitDir.Get(region + '/' + proc), -1.)
else:
postfitSub = None
plotConfig = getConfig(region)
lumi = 0.
for sample in plotConfig.obs.samples:
lumi += sample.lumi / plotConfig.prescales[sample]
canvas = RatioCanvas(lumi = lumi, name = region)
canvas.legend.setPosition(0.6, 0.6, 0.9, 0.9)
canvas.legend.add('obs', title = 'Observed', opt = 'LP', color = ROOT.kBlack, mstyle = 8, msize = 0.8)
canvas.legend.add('prefit', title = 'Prefit total', opt = 'LF', color = ROOT.kRed, lstyle = ROOT.kDashed, lwidth = 2, fstyle = 3004, mstyle = 8, msize = 0.8)
if postfitSub:
canvas.legend.add('subdom', title = 'Postfit subdominant', opt = 'F', fcolor = ROOT.kGray, fstyle = 1001)
canvas.legend.add('postfit', title = 'Postfit total', opt = 'LF', color = ROOT.kBlue, lstyle = ROOT.kSolid, lwidth = 2, fstyle = 3005, mstyle = 8, msize = 0.8)
obs.SetTitle('')
prefitTotal.SetTitle('')
postfitTotal.SetTitle('')
if postfitSub:
postfitSub.SetTitle('')
canvas.legend.apply('obs', obs)
canvas.legend.apply('prefit', prefitTotal, opt = 'L')
lumi = 0.
for sample in plotConfig.obs.samples:
lumi += sample.lumi / plotConfig.prescales[sample]
if not SIMPLE:
printRegionHeader(region, prefitTotal)
if pdir == 's':
pftitle = 'Sig.+bkgd. fit'
elif pdir == 'c':
pftitle = 'CR-only fit'
else:
pftitle = 'Bkgd. fit'
canvas = RatioCanvas(lumi=lumi, name=region, prelim=False)
canvas.legend.ncolumns = 1
# y coordinates will be adjusted later
resultLegend = Legend(0.35, 0., 0.7, 1.)
resultLegend.add('obs',
title='Data',
opt='ELP',
color=ROOT.kBlack,
mstyle=8,
msize=0.8)
resultLegend.add('prefit',
title='Pre-fit',
opt='LF',
color=ROOT.kRed,
lstyle=ROOT.kDashed,
