class FakeRateCalculator(object):
def __init__(self, looseFilesData, tightFilesData,
looseFilesMC, tightFilesMC, channels, outFile,
intLumi=-1., fakeFactor=False):
if looseFilesData:
assert tightFilesData, "I can only make a fake rate from data with both a tight and loose ntuple"
self.dataFiles = {'num' : tightFilesData, 'denom' : looseFilesData}
else:
assert len(tightFilesData)==0, "I can only make a fake rate from data with both a tight and loose ntuple"
self.dataFiles = {}
if looseFilesMC:
assert tightFilesMC, "I can only make a fake rate from Monte Carlo with both a tight and loose ntuple"
self.mcFiles = {'numMC' : tightFilesMC, 'denomMC' : looseFilesMC}
else:
assert len(tightFilesMC)==0, "I can only make a fake rate from Monte Carlo with both a tight and loose ntuple"
self.mcFiles = {}
self.hasData = bool(self.dataFiles)
self.hasMC = bool(self.mcFiles)
self.outFileName = outFile
outFileBaseName = self.outFileName.split('/')[-1]
self.outDir = self.outFileName.replace(outFileBaseName, '')
self.intLumi = intLumi
self.plotter = NtuplePlotter(channels, self.outDir,
self.mcFiles, self.dataFiles,
self.intLumi)
self.channels = parseChannels(channels)
assert len(self.channels) > 0 and \
all(len(ch) == 3 for ch in self.channels), \
"Invalid channel %s. Channels must have 3 objects."%channels
self.WrappedHists = [
self.plotter.WrappedHist,
self.plotter.WrappedHist2,
self.plotter.WrappedHist3,
]
self.WrappedStacks = [
self.plotter.WrappedStack,
self.plotter.WrappedStack,
self.plotter.WrappedHist3, # THistStack only goes up to 2-D
]
self.histMakers = [
lambda *args, **kwargs: self.plotter.makeHist(*args,
perUnitWidth=False,
**kwargs),
self.plotter.makeHist2,
self.plotter.makeHist3,
]
self.stackMakers = [
lambda *args, **kwargs: self.plotter.makeStack(*args,
perUnitWidth=False,
**kwargs),
self.plotter.makeStack2, # THistStack same for 1- and 2-D
self.plotter.makeStack3,
]
self.extractHistFromStack = [
lambda s: self.WrappedHists[0](asrootpy(s.GetStack().Last()),
category=s.getCategory(),
variable=s.getVariable(),
selection=s.getSelection()),
lambda s: self.WrappedHists[1](asrootpy(s.GetStack().Last()),
category=s.getCategory(),
variable=s.getVariable(),
selection=s.getSelection()),
lambda s: s, # THistStack only goes up to 2-D
]
self.outputs = []
self.fakeFactor = fakeFactor
def calculateFakeRate(self, name, channels, *varsAndBinnings,
**kwargs):
'''
Make a fake rate histogram. It is stored in self.outputs and returned.
Channel should be one channel or a list of channels. Correct object to
use for each channel is figured out automatically.
The next 2, 4, or 6 arguments should be one, two, or three variables
to bin in, each followed immediately by an iterable containing
bin edges or [nbins, min, max].
If keyword argument 'useMC' evaluates to True, MC samples are used
instead of data.
If keyword argument 'draw' evaluates to True, a .png file is created
in the same directory as the eventual output with a plot of the
fake rate.
If a list of subtraction samples is provided, by kwargs['subtractSamples'],
the MC contributions of these samples are subtracted from the numerator and
denominator before the fake rate is calculated. No bin may be less
than 0.
'''
subtractSamples = kwargs.pop('subtractSamples', [])
if len(varsAndBinnings) % 2 == 1 or len(varsAndBinnings) < 2 or len(varsAndBinnings) > 6:
raise ValueError("Invalid list of variables and their binnings")
nDims = len(varsAndBinnings) / 2
varTemplate = ''
binning = []
varTemplate = ':'.join("{0}"+varsAndBinnings[2*i] for i in xrange(nDims))
for i in xrange(nDims):
ind = 2*i+1
if len(varsAndBinnings[ind]) == 3 and isinstance(varsAndBinnings[ind][0], int):
binning += varsAndBinnings[ind] # uniform bins
else:
binning.append(varsAndBinnings[ind]) # variable bins
if isinstance(channels, str):
channels = [channels]
varList = [varTemplate.format(self.fakeObjectForChannel(ch)) for ch in channels]
selecList = ["" for v in varList]
outputs = []
drawablesMC = {}
if self.hasMC:
samplesToDraw = [s for s in self.plotter.ntuples['numMC'].keys() if s not in subtractSamples]
sNum = self.stackMakers[nDims-1]("numMC", samplesToDraw, channels,
varList, selecList, binning,
weight="GenWeight")
drawablesMC['num'] = sNum
numMC = self.extractHistFromStack[nDims-1](sNum)
sDenom = self.stackMakers[nDims-1]("denomMC", samplesToDraw,
channels, varList,
selecList, binning,
weight="GenWeight")
drawablesMC['denom'] = sDenom
denomMC = self.extractHistFromStack[nDims-1](sDenom)
fMC = self.WrappedHists[nDims-1](asrootpy(numMC.Clone()),
name=name+"MC_fakeRate",
isData=False)
fMC.Divide(denomMC)
drawablesMC['fakeRate'] = fMC
if self.fakeFactor:
# actual scale factor, f/(1-f)
outMC = self.WrappedHists[nDims-1](asrootpy(numMC.empty_clone()),
name=name+"MC")
for fb, b in zip(fMC, outMC):
if b.overflow:
continue
b.value = fb.value / (1. - fb.value)
else:
outMC = self.WrappedHists[nDims-1](fMC.clone(), name=name+"MC")
for b in outMC:
if b.overflow:
b.value = 0.
b.error = 0.
outputs.append(outMC)
drawablesData = {}
if self.hasData:
num = self.histMakers[nDims-1]("num", "num", channels, varList,
selecList, binning)
denom = self.histMakers[nDims-1]("denom", "denom", channels, varList,
selecList, binning)
num.sumw2()
denom.sumw2()
for ss in subtractSamples:
subNum = self.histMakers[nDims-1]("numMC", ss, channels,
varList, selecList, binning,
1., weights="GenWeight")
num -= subNum
subDenom = self.histMakers[nDims-1]("denomMC", ss, channels,
varList, selecList, binning,
1., weights="GenWeight")
denom -= subDenom
for bNum, bDenom in zip(num, denom):
if bNum.value < 0. or bDenom.value <= 0.:
bNum.value = 0.
bNum.error = 0.
bDenom.value = 0.0000001
bDenom.error = 0.
num.sumw2()
denom.sumw2()
f = self.WrappedHists[nDims-1](asrootpy(num.Clone()),
name=name+"_fakeRate")
f.Divide(denom)
drawablesData['num'] = num
drawablesData['denom'] = denom
drawablesData['fakeRate'] = f
if self.fakeFactor:
out = self.WrappedHists[nDims-1](asrootpy(num.empty_clone()),
name=name)
for fb, b in zip(f, out):
if b.overflow:
continue
b.value = fb.value / (1. - fb.value)
else:
out = self.WrappedHists[nDims-1](f.clone(), name=name)
for b in out:
if b.overflow:
b.value = 0.
b.error = 0.
outputs.append(out)
for o in outputs:
for i in xrange(nDims):
o.axis(i).title = varsAndBinnings[i*2]
if kwargs.pop('draw', False):
self.drawFakeRate(name, nDims, drawablesData, drawablesMC, **kwargs)
self.outputs += outputs
return outputs
def drawFakeRate(self, name, nDims, drawablesData, drawablesMC, **kwargs):
'''
Create a .png with the fake rate plot, with numbers superimposed over
each (1- and 2-D only), and for the numerator and denominator. For 1-D,
MC and data are drawn on the same plot.
'''
drawings = {}
if nDims == 1 and bool(drawablesData) and bool(drawablesMC):
for typeToPlot in ['num', 'denom', 'fakeRate']:
drawings[typeToPlot] = self.plotter.Drawing(name+'_'+typeToPlot,
self.plotter.style,
1000, 1000)
if typeToPlot == 'fakeRate':
drawablesMC[typeToPlot].drawstyle = 'hist'
drawablesMC[typeToPlot].color = 'red'
drawings[typeToPlot].addObject(drawablesMC[typeToPlot],
legendName='MC',
legendStyle="L")
else:
drawings[typeToPlot].addObject(drawablesMC[typeToPlot])
# drawings[typeToPlot].addRatio(drawablesData[typeToPlot],
# drawablesMC[typeToPlot], 0.23,
# yTitle="Data / MC")
if typeToPlot == 'fakeRate':
fakeRateData = self.plotter.WrappedGraph(type='asymm',
title='data')
fakeRateData.Divide(drawablesData['num'], drawablesData['denom'])
else:
fakeRateData = drawablesData[typeToPlot]
fakeRateData.color = 'black'
fakeRateData.drawstyle = 'PE'
drawings[typeToPlot].addObject(fakeRateData,
legendName='data',
legendStyle='LPE')
drawings[typeToPlot].draw({'yerror_in_padding' : False},
'%s/%s_%s.png'%(self.outDir, name,
typeToPlot),
xTitle=kwargs.get('xTitle', ''),
xUnits=kwargs.get('xUnits', ''),
yTitle='Fake Rate',
intLumi=self.intLumi,
perUnitWidth=False)
else:
for maybeMC in ['', 'MC']:
if maybeMC:
drawables = drawablesMC
else:
drawables = drawablesData
if not drawables:
continue
for typeToPlot, plot in drawables.iteritems():
ttp = typeToPlot+maybeMC
drawings[ttp] = self.plotter.Drawing(name+'_'+ttp,
self.plotter.style,
800, 800)
if isinstance(plot, _Hist2D):
plot.drawstyle = plot.drawstyle+"TEXT"
if "COLZ" not in plot.drawstyle:
plot.drawstyle = plot.drawstyle+"COLZ"
if isinstance(plot, _Hist) and maybeMC:
plot.drawstyle='hist'
plot.color='red'
drawings[ttp].addObject(plot,
addToLegend=(bool(maybeMC) and typeToPlot != 'fakeRate' and nDims == 1),
legendName=("MC" if maybeMC else "data"),
legendStyle=("F" if maybeMC else "LPE"))
drawings[ttp].draw({'yerror_in_padding' : False},
'%s/%s_%s.png'%(self.outDir,
name, ttp),
xTitle=kwargs.get('xTitle', ''),
xUnits=kwargs.get('xUnits', ''),
intLumi=self.intLumi,
stackErr=(nDims==1),
perUnitWidth=False)
_fakeRateCalculator_channelObjMap_ = {}
def fakeObjectForChannel(self, channel):
'''
Get the odd object out for this channel (the l in Z+l).
E.g. 'm3' for channel 'mmm' or 'e' for channel 'emm'.
'''
try:
return self._fakeRateCalculator_channelObjMap_[channel]
except KeyError:
assert len(channel) == 3 and any(channel.count(ob)>1 for ob in channel),\
"Only Z+l-like channels are allowed, %s will not work!"%channel
objects = mapObjects(channel)
for obj in objects:
if len(obj) == 1: # only object of its type
self._fakeRateCalculator_channelObjMap_[channel] = obj
return obj
if obj[-1] == '3': # Z always listed first for lll channels
self._fakeRateCalculator_channelObjMap_[channel] = obj
return obj
else: # shouldn't ever happen
raise
def writeOutput(self):
with root_open(self.outFileName, "RECREATE") as f:
for op in self.outputs:
op.Write()
def __init__(self, looseFilesData, tightFilesData,
looseFilesMC, tightFilesMC, channels, outFile,
intLumi=-1., fakeFactor=False):
if looseFilesData:
assert tightFilesData, "I can only make a fake rate from data with both a tight and loose ntuple"
self.dataFiles = {'num' : tightFilesData, 'denom' : looseFilesData}
else:
assert len(tightFilesData)==0, "I can only make a fake rate from data with both a tight and loose ntuple"
self.dataFiles = {}
if looseFilesMC:
assert tightFilesMC, "I can only make a fake rate from Monte Carlo with both a tight and loose ntuple"
self.mcFiles = {'numMC' : tightFilesMC, 'denomMC' : looseFilesMC}
else:
assert len(tightFilesMC)==0, "I can only make a fake rate from Monte Carlo with both a tight and loose ntuple"
self.mcFiles = {}
self.hasData = bool(self.dataFiles)
self.hasMC = bool(self.mcFiles)
self.outFileName = outFile
outFileBaseName = self.outFileName.split('/')[-1]
self.outDir = self.outFileName.replace(outFileBaseName, '')
self.intLumi = intLumi
self.plotter = NtuplePlotter(channels, self.outDir,
self.mcFiles, self.dataFiles,
self.intLumi)
self.channels = parseChannels(channels)
assert len(self.channels) > 0 and \
all(len(ch) == 3 for ch in self.channels), \
"Invalid channel %s. Channels must have 3 objects."%channels
self.WrappedHists = [
self.plotter.WrappedHist,
self.plotter.WrappedHist2,
self.plotter.WrappedHist3,
]
self.WrappedStacks = [
self.plotter.WrappedStack,
self.plotter.WrappedStack,
self.plotter.WrappedHist3, # THistStack only goes up to 2-D
]
self.histMakers = [
lambda *args, **kwargs: self.plotter.makeHist(*args,
perUnitWidth=False,
**kwargs),
self.plotter.makeHist2,
self.plotter.makeHist3,
]
self.stackMakers = [
lambda *args, **kwargs: self.plotter.makeStack(*args,
perUnitWidth=False,
**kwargs),
self.plotter.makeStack2, # THistStack same for 1- and 2-D
self.plotter.makeStack3,
]
self.extractHistFromStack = [
lambda s: self.WrappedHists[0](asrootpy(s.GetStack().Last()),
category=s.getCategory(),
variable=s.getVariable(),
selection=s.getSelection()),
lambda s: self.WrappedHists[1](asrootpy(s.GetStack().Last()),
category=s.getCategory(),
variable=s.getVariable(),
selection=s.getSelection()),
lambda s: s, # THistStack only goes up to 2-D
]
self.outputs = []
self.fakeFactor = fakeFactor
mcSamples = {
'mc':'/data/nawoods/ntuples/zzNtuples_mc_{1}/results_{0}/ZZTo4L_13TeV_*.root,/data/nawoods/ntuples/zzNtuples_mc_{1}/results_{0}/GluGlu*.root'.format(sampleID, ntupleSet),
'mc3P1F':'/data/nawoods/ntuples/zzNtuples_mc_{1}/results_{0}_3P1F/*.root'.format(sampleID, ntupleSet3P1F),
'mc2P2F':'/data/nawoods/ntuples/zzNtuples_mc_{1}/results_{0}_2P2F/*.root'.format(sampleID, ntupleSet2P2F),
}
if args.dySkim:
mcSamples['mc'] += ',/data/nawoods/ntuples/zzNtuples_mc_dySkim_10feb2016_0/results_{0}/DYSkim_*.root'.format(sampleID, ntupleSet)
dataSamples = {
'data':'/data/nawoods/ntuples/zzNtuples_data_2015silver_{1}/results_{0}/data*.root'.format(sampleID, ntupleSet),
'3P1F':'/data/nawoods/ntuples/zzNtuples_data_2015silver_{1}/results_{0}_3P1F/data*.root'.format(sampleID, ntupleSet3P1F),
'2P2F':'/data/nawoods/ntuples/zzNtuples_data_2015silver_{1}/results_{0}_2P2F/data*.root'.format(sampleID, ntupleSet2P2F),
}
plotter = NtuplePlotter('zz', '/afs/cern.ch/user/n/nawoods/www/ZZPlots/counting_{0}_{1}'.format(date.today().strftime('%d%b%Y').lower(), ana),
mcSamples, dataSamples,
intLumi=2619.)
basicSelection = ''
print ""
if args.printData:
print "Signal (data):"
plotter.printPassingEvents('data')
elif not args.eventsOnly:
print "Signal (data):"
print ' eeee: %d'%plotter.ntuples['data']['data']['eeee'].GetEntries()
print ' eemm: %d'%plotter.ntuples['data']['data']['eemm'].GetEntries()
print ' mmmm: %d'%plotter.ntuples['data']['data']['mmmm'].GetEntries()
print ""
if args.print3P1F: