def getPlotsIn(inF, dirname, LPdirname, rebin=True, var=None, bins=[]):
"""gets the data and mc sums in a given directory"""
data, dataLP = None, None
if not inF.Get(LPdirname): return None, None
LPKeys = inF.Get(LPdirname).GetListOfKeys()
Keys = inF.Get(dirname).GetListOfKeys()
for key, LPkey in product(Keys, LPKeys):
name = key.GetName()
nameLP = LPkey.GetName()
if ('Graph' in name): continue
if ('Graph' in nameLP): continue
if (not name == dirname) and (not nameLP == LPdirname): continue
#data histogram
data_ = key.ReadObj()
dataLP_ = LPkey.ReadObj()
# if data_.InheritsFrom(ROOT.TH2.Class()): continue
# if dataLP_.InheritsFrom(ROOT.TH2.Class()): continue
if rebin:
data_.Rebin()
dataLP_.Rebin()
if ((not bins == []) and (var in name)):
data = rebinUnequalBinSize(data_, bins)
dataLP = rebinUnequalBinSize(dataLP_, bins)
else:
data = data_.Clone()
data.SetDirectory(0)
fixExtremities(data)
dataLP = dataLP_.Clone()
dataLP.SetDirectory(0)
fixExtremities(dataLP)
return data, dataLP
def customizeSignalShapes(binName, sigName, sigF, baseSigF, outDir):
"""
takes specific signal and scales the base signal simulation + systematics by the ratio sig/orig_signal
the result is stored in {0}_{1}.shapes.root {0}=signal name {1}=binName
"""
#get signal
fIn = ROOT.TFile.Open(sigF)
sigH = fIn.Get('{0}_mlb'.format(binName))
sigH.SetDirectory(0)
fixExtremities(
sigH
) # this needs to be done as the signal is the raw output of TOP-17-010.cc
fIn.Close()
#get original signal
fIn = ROOT.TFile.Open(baseSigF)
origSigH = fIn.Get('{0}_mlb/central'.format(binName))
origSigH.SetDirectory(0)
#force the integral to be the same as this is a shape analysis
sf = origSigH.Integral() / sigH.Integral()
sigH.Scale(sf)
#transfer factor
tfH = sigH.Clone('tf')
tfH.Divide(origSigH)
tfH.SetDirectory(0)
#prepare the output applying the transfer factor to each histogram found
fOut = ROOT.TFile.Open(
os.path.join(outDir, '{0}.shapes.root'.format(sigName)), 'RECREATE')
fdir = fOut.mkdir('{0}_mlb'.format(binName))
for k in fIn.Get('{0}_mlb'.format(binName)).GetListOfKeys():
h = k.ReadObj()
h.SetDirectory(fdir)
for xbin in range(h.GetNbinsX()):
h.SetBinContent(
xbin + 1,
h.GetBinContent(xbin + 1) * tfH.GetBinContent(xbin + 1))
fdir.cd()
h.Write()
fOut.Close()
#free mem
tfH.Delete()
fIn.Close()
def getPassGenDistributions(pName, gName, plotter, ci, fill=0, marker=20):
"""reads the distributions from the plotter"""
passH = plotter.Get(pName)
fixExtremities(passH)
passH.SetDirectory(0)
passH.SetFillStyle(fill)
passH.SetLineColor(ci)
passH.SetMarkerColor(ci)
passH.SetLineWidth(2)
passH.SetMarkerStyle(marker)
genH = plotter.Get(gName)
fixExtremities(genH)
genH.SetDirectory(0)
genH.SetFillStyle(fill)
genH.SetLineColor(ci)
genH.SetMarkerColor(ci)
genH.SetLineWidth(2)
genH.SetMarkerStyle(marker)
return passH, genH
def getPlotsIn(inF,dirname,specList=[],rebin=True,bins=[]):
"""gets the data and mc sums in a given directory"""
data,mcTotal,h=None,None,None
mcSpecList=[None]*len(specList)
for key in inF.Get(dirname).GetListOfKeys():
name=key.GetName()
if 'Graph' in name : continue
#data histogram
if name==dirname:
data_=key.ReadObj()
if rebin :
data_.Rebin(rebin)
if len(bins)>0:
data = rebinUnequalBinSize(data_,bins)
else:
data = data_.Clone()
data.SetDirectory(0)
fixExtremities(data)
#predictions
else:
h_=key.ReadObj()
if rebin:
h_.Rebin(rebin)
if len(bins)>0:
h = rebinUnequalBinSize(h_,bins)
else:
h = h_.Clone();
proc=name.split('_')[-1]
#check if this specific process should be stored
try:
idx=specList.index(proc)
mcSpecList[idx]=h.Clone()
mcSpecList[idx].SetDirectory(0)
fixExtremities(mcSpecList[idx])
except:
pass
#add to the total
if not mcTotal:
mcTotal=h.Clone(dirname+'_mctot')
mcTotal.SetDirectory(0)
mcTotal.Reset('ICE')
mcTotal.Add(h)
if mcTotal : fixExtremities(mcTotal)
return data,mcTotal,mcSpecList
def getUncertaintiesFromProjection(opt,fIn,d,proc_systs,hnom):
"""projects the _exp and _th 2D histograms to build the corresponding Up/Down uncertainty templates"""
histos=[]
errors=[]
warns=[]
#map all systematics available for projection
allSysts={}
for hname in ['{0}_exp'.format(d),'{0}_th'.format(d)]:
h2d=fIn.Get('{0}/{0}_{1}'.format(hname,proc_systs['title']))
try:
for ybin in range(h2d.GetNbinsY()):
allSysts[h2d.GetYaxis().GetBinLabel(ybin+1)]=(h2d,ybin+1)
except:
pass
#project systematics
for s in proc_systs['proj']:
slist,norm,doEnvelope,smooth=proc_systs['proj'][s]
try:
h2d,ybin=allSysts[ slist[0] ]
varUp=h2d.ProjectionX('varup',ybin,ybin)
fixExtremities(varUp) #add the overflow as for the main plot
if smooth: applySmoothing(varUp)
except:
errors.append('Failed to prepare %s'%slist[0])
continue
varDn=None
#project down variation if available
if not doEnvelope:
try:
h2d,ybin=allSysts[ slist[1] ]
varDn=h2d.ProjectionX('vardn',ybin,ybin)
fixExtremities(varDn) #add the overflow as for the main plot
if smooth: applySmoothing(varDn)
except:
errors.append('Failed to prepare %s'%slist[1])
continue
#make an envelope if several are available
elif len(slist)>2:
try:
varUp.Reset('ICE')
varUp.Add(hnom)
varDn=hnom.Clone('vardn')
for s_i in slist:
h2d,ybin=allSysts[ s_i ]
vartemp=h2d.ProjectionX('vartemp',ybin,ybin)
fixExtremities(vartemp) #add the overflow as for the main plot
if smooth: applySmoothing(vartemp)
#do max(var_i-nom,var_j-nom) per bin
for xbin in range(hnom.GetNbinsX()):
val = vartemp.GetBinContent(xbin+1)
deltaVal = val-hnom.GetBinContent(xbin+1)
if deltaVal>0:
curDeltaVal = varUp.GetBinContent(xbin+1)-hnom.GetBinContent(xbin+1)
if curDeltaVal<deltaVal:
varUp.SetBinContent(xbin+1,val)
else:
curDeltaVal = varDn.GetBinContent(xbin+1)-hnom.GetBinContent(xbin+1)
if curDeltaVal>deltaVal:
varDn.SetBinContent(xbin+1,val)
#delete as no longer used
vartemp.Delete()
except:
errors.append('Failed to prepare %s'%s_i)
continue
#mirror the shape if down variation is not available yet
if not varDn:
varDn=getMirrored(varUp,hnom,'vardn')
#check, format and add to the list if variation is to keep
keep,checkReport=doFinalTemplateCheck(hnom,varUp,varDn)
if len(checkReport):
pfix='Keep but' if keep else 'Discard as'
warns.append('%s %s for %s'%(pfix,checkReport.replace('\n',','),s))
formatTemplate(varUp,s+'Up',norm=hnom.Integral() if norm else None)
formatTemplate(varDn,s+'Down',norm=hnom.Integral() if norm else None)
if keep:
histos.append(varUp)
histos.append(varDn)
#show plot with warnings found
if opt.debug:
showVariation(hnom,
[varUp,varDn],
checkReport.split('\n'),
os.path.join(opt.output,'{0}_{1}_{2}'.format(hnom.GetTitle(),d,s)))
#print errors/warnings found
if len(errors) or len(warns):
print '-'*50
print 'Some errors/warnings found projecting syst templates for',proc_systs['title']
print 'Please check the list below'
for e in errors : print e
for w in warns : print w
print '-'*50
return histos
def customizeData(binName, dataDef, bkgList, templDir, outDir):
""" customizes data to use as observation """
dataDef = dataDef.replace("\"", "")
tkns = dataDef.split(',')
dataType, dataF, dataHisto = tkns[0:3]
#get the data
inF = ROOT.TFile.Open(dataF)
dataH = inF.Get(dataHisto)
dataH.SetDirectory(0)
inF.Close()
#use a scenario as pseudo-data
if len(tkns) == 5:
#reset current histogram
ndata = dataH.Integral()
dataH.Reset('ICE')
#scale scenario to the current number of entries
scenario, scenarioHisto = tkns[3:5]
scenarioF = os.path.dirname(os.path.dirname(dataF))
scenarioF = os.path.join(scenarioF, scenario)
print 'Using a scenario as pseudo-data', scenarioF, scenarioHisto
inF = ROOT.TFile.Open(scenarioF)
hscen = inF.Get(scenarioHisto)
fixExtremities(
hscen
) # this needs to be done as the signal is the raw output of TOP-17-010.cc
hscen.Scale(ndata / hscen.Integral())
dataH.Add(hscen)
inF.Close()
dataH.SetTitle(dataH.GetTitle() + os.path.dirname(scenario))
print 'Scaled', scenario, 'to', ndata, 'new title=', dataH.GetTitle()
#specific for pseudo-data
#sum up signal to background expectations when dataType is 'sig' type
#round each bin to an integer
if dataType == 'sig':
for proc in bkgList:
inF = ROOT.TFile.Open('{0}/templates_{1}.root'.format(
templDir, proc))
h = inF.Get('{0}_mlb/central'.format(binName))
dataH.Add(h)
inF.Close()
for xbin in range(dataH.GetNbinsX()):
dataH.SetBinContent(xbin + 1, int(dataH.GetBinContent(xbin + 1)))
dataType = dataH.GetTitle()
for tkn in [' ', '#', '_', '^', '{', '}', ',', ':', '+', '-']:
dataType = dataType.replace(tkn, '')
dataType = dataType.replace('.', 'p')
dataType = 'pseudodata.%s' % dataType
#save to file
dataShapesURL = os.path.join(outDir, '%s.shapes.root' % dataType)
fOut = ROOT.TFile.Open(dataShapesURL, 'RECREATE')
fdir = fOut.mkdir('{0}_mlb'.format(binName))
fdir.cd()
dataH.SetTitle('')
dataH.SetDirectory(fdir)
dataH.Write('data_obs')
fOut.Close()
return dataShapesURL
def computeVBFTriggerEff(f, distsOfInterest):
"""computes ratios to evaluate VBF trigger efficiency and fits an erf function"""
ratios = {}
inF = ROOT.TFile.Open(f)
for key in inF.GetListOfKeys():
name = key.GetName()
if not 'HighPtVBFA' in name: continue
dist = '_'.join(name.split('_')[1:])
if not dist in distsOfInterest: continue
dataNum, mcNum, _ = getPlotsIn(inF, name, rebin=False)
dataDen, mcDen, _ = getPlotsIn(inF,
name.replace('HighPtVBFA',
'HighPtOfflineVBFA'),
rebin=False)
for h in [dataNum, mcNum, dataDen, mcDen]:
fixExtremities(h)
scaleTo(h, 1.0)
dataNum.Divide(dataDen)
dataNum.GetYaxis().SetTitle(
'High p_{T} VBF #gamma / High p_{T} #gamma')
dataNum.SetTitle('Data')
dataNum.SetMarkerStyle(20)
mcNum.Divide(mcDen)
mcNum.GetYaxis().SetTitle('High p_{T} VBF #gamma / High p_{T} #gamma')
mcNum.SetTitle('MC')
mcNum.SetMarkerStyle(24)
mcNum.SetMarkerColor(ROOT.kGray)
mcNum.SetLineColor(ROOT.kGray)
mcNum.SetFillStyle(1001)
mcNum.SetFillColor(ROOT.kGray)
data2mc = dataNum.Clone('{0}_data2mc'.format(dist))
data2mc.Divide(mcNum)
data2mc.SetDirectory(0)
data2mc.GetYaxis().SetTitle('Data/MC')
turnOn = None
if dist in ['mjj', 'subleadpt']:
turnOn = ROOT.TF1('turnon', '[0]+[1]/(1.+TMath::Exp([2]*(x-[3])))',
dataNum.GetXaxis().GetXmin(),
dataNum.GetXaxis().GetXmax())
turnOn.SetParLimits(0, 0., 0.8)
turnOn.SetParLimits(1, 0.5, 1.2)
turnOn.SetParLimits(2, -2, 2)
turnOn.SetParLimits(3, 500, 2000)
if dist == 'subleadpt': turnOn.SetParLimits(3, 20, 50)
#else:
# turnOn=ROOT.TF1('turnon',
# '[0]',
# dataNum.GetXaxis().GetXmin(),
# dataNum.GetXaxis().GetXmax())
data2mcParam = parametrizeTurnOn(dataNum, mcNum, turnOn)
ratios[name] = (dataNum, mcNum, data2mc, data2mcParam)
inF.Close()
return ratios