def initCache(self, cacheDir):
if cacheDir:
self.cacheDir = cacheDir
cacheFileName = os.path.join(cacheDir, self.name + '.pkl')
if not os.path.exists(os.path.dirname(cacheFileName)):
os.makedirs(os.path.dirname(cacheFileName))
self.cache = Cache(cacheFileName, verbosity=2)
else:
self.cache = None
def __init__(self, name, cacheDir=None, useTop16009=False):
super(DataDrivenTTZEstimate, self).__init__(name, cacheDir=cacheDir)
self.nJets = (3, -1) # jet selection (min, max)
self.nLooseBTags = (2, -1) # loose bjet selection (min, max)
self.nMediumBTags = (0, -1) # bjet selection (min, max)
self.useTop16009 = useTop16009
self.ratioTop16009 = 1.27 #
self.sysErrTop16009 = (-0.17, +0.20)
self.statErrTop16009 = (-0.37, +0.42)
# Because we are going to reuse a lot of yields which otherwise will be terribly slow
self.helperCacheName = os.path.join('.', 'helperCache.pkl')
self.helperCache = Cache(self.helperCacheName, verbosity=2)
def initCache(self, cacheDir):
if cacheDir:
self.cacheDir=cacheDir
cacheFileName = os.path.join(cacheDir, self.name+'.pkl')
if not os.path.exists(os.path.dirname(cacheFileName)):
os.makedirs(os.path.dirname(cacheFileName))
self.cache = Cache(cacheFileName, verbosity=2)
else:
self.cache=None
def __init__(self, name, cacheDir=None, useTop16009=False):
super(DataDrivenTTZEstimate, self).__init__(name, cacheDir=cacheDir)
self.nJets = (3,-1) # jet selection (min, max)
self.nLooseBTags = (2,-1) # loose bjet selection (min, max)
self.nMediumBTags = (0,-1) # bjet selection (min, max)
self.useTop16009 = useTop16009
self.ratioTop16009 = 1.27 #
self.sysErrTop16009 = (-0.17, +0.20)
self.statErrTop16009 = (-0.37, +0.42)
# Because we are going to reuse a lot of yields which otherwise will be terribly slow
self.helperCacheName = os.path.join('.', 'helperCache.pkl')
self.helperCache = Cache(self.helperCacheName, verbosity=2)
from StopsDilepton.analysis.u_float import u_float
from math import sqrt
##https://twiki.cern.ch/twiki/bin/viewauth/CMS/SUSYSignalSystematicsRun2
from StopsDilepton.tools.user import combineReleaseLocation
from StopsDilepton.tools.cardFileWriter import cardFileWriter
limitPrefix = options.regions
limitDir = os.path.join(setup.analysis_results, setup.prefix(), 'cardFiles',
limitPrefix)
overWrite = False
useCache = True
verbose = True
if not os.path.exists(limitDir): os.makedirs(limitDir)
cacheFileName = os.path.join(limitDir, 'calculatedLimits.pkl')
limitCache = Cache(cacheFileName, verbosity=2)
def wrapper(s):
c = cardFileWriter.cardFileWriter()
c.releaseLocation = combineReleaseLocation
cardFileName = os.path.join(limitDir, s.name + '.txt')
if not os.path.exists(cardFileName) or overWrite:
counter = 0
c.reset()
c.addUncertainty('PU', 'lnN')
c.addUncertainty('topPt', 'lnN')
c.addUncertainty('JEC', 'lnN')
c.addUncertainty('JER', 'lnN')
c.addUncertainty('SFb', 'lnN')
class DataDrivenTTZEstimate(SystematicEstimator):
def __init__(self, name, cacheDir=None, useTop16009=False):
super(DataDrivenTTZEstimate, self).__init__(name, cacheDir=cacheDir)
self.nJets = (3, -1) # jet selection (min, max)
self.nLooseBTags = (2, -1) # loose bjet selection (min, max)
self.nMediumBTags = (0, -1) # bjet selection (min, max)
self.useTop16009 = useTop16009
self.ratioTop16009 = 1.27 #
self.sysErrTop16009 = (-0.17, +0.20)
self.statErrTop16009 = (-0.37, +0.42)
# Because we are going to reuse a lot of yields which otherwise will be terribly slow
self.helperCacheName = os.path.join('.', 'helperCache.pkl')
self.helperCache = Cache(self.helperCacheName, verbosity=2)
def yieldFromCache(self, setup, sample, channel, selectionString,
weightString):
s = (sample, channel, selectionString, weightString)
if self.helperCache.contains(s):
return self.helperCache.get(s)
else:
yieldFromDraw = u_float(
**setup.sample[sample][channel].getYieldFromDraw(
selectionString, weightString))
self.helperCache.add(s, yieldFromDraw, save=True)
return yieldFromDraw
#Concrete implementation of abstract method 'estimate' as defined in Systematic
def _estimate(self, region, channel, setup):
logger.info("Data-driven TTZ estimate for region " + str(region) +
" in channel " + channel + " and setup " + str(setup.sys) +
":")
#Sum of all channels for 'all'
if channel == 'all':
estimate = sum([
self.cachedEstimate(region, c, setup)
for c in ['MuMu', 'EE', 'EMu']
])
else:
zWindow = 'allZ' if channel == 'EMu' else 'offZ'
preSelection = setup.preselection('MC',
zWindow=zWindow,
channel=channel)
MC_2l = "&&".join([
region.cutString(setup.sys['selectionModifier']),
preSelection['cut']
])
weight = setup.weightString()
logger.info("weight: %s", weight)
yield_ttZ_2l = setup.lumi[channel] / 1000. * self.yieldFromCache(
setup, 'TTZ', channel, MC_2l, weight)
logger.info("yield_MC_2l: %s" % yield_ttZ_2l)
if self.useTop16009:
sysError = max((abs(x) for x in self.sysErrTop16009
)) # not sure yet to handle assymetric errors
statError = max((abs(x) for x in self.statErrTop16009))
error = sqrt(sysError * sysError + statError * statError)
return u_float(self.ratioTop16009, error) * yield_ttZ_2l
else:
# pt leptons > 30, 20, 10 GeV
useTrigger = False # setup.parameters['useTriggers'] # better not to use three lepton triggers, seems to be too inefficient
lllSelection = {}
lllSelection['MuMu'] = "&&".join([
getLeptonString(3, 0),
getPtThresholdString(30, 20, 10)
]) + ("&&HLT_3mu" if useTrigger else "")
lllSelection['MuMuE'] = "&&".join([
getLeptonString(2, 1),
getPtThresholdString(30, 20, 10)
]) + ("&&HLT_2mu1e" if useTrigger else "")
lllSelection['MuEE'] = "&&".join([
getLeptonString(1, 2),
getPtThresholdString(30, 20, 10)
]) + ("&&HLT_2e1mu" if useTrigger else "")
lllSelection['EE'] = "&&".join([
getLeptonString(0, 3),
getPtThresholdString(30, 20, 10)
]) + ("&&HLT_3e" if useTrigger else "")
lllSelection['EMu'] = "((" + lllSelection[
'MuMuE'] + ")||(" + lllSelection['MuEE'] + "))"
bJetSelectionM = "(Sum$(JetGood_pt>30&&abs(JetGood_eta)<2.4&&JetGood_id&&JetGood_btagCSV>0.890))"
bJetSelectionL = "(Sum$(JetGood_pt>30&&abs(JetGood_eta)<2.4&&JetGood_id&&JetGood_btagCSV>0.605))"
zMassSelection = "abs(mlmZ_mass-91.1876)<10"
# Start from base hadronic selection and add loose b-tag and Z-mass requirement
selection = {}
for dataOrMC in ["Data", "MC"]:
selection[dataOrMC] = setup.selection(
dataOrMC,
hadronicSelection=True,
**setup.defaultParameters(
update={
'nJets': self.nJets,
'nBTags': self.nMediumBTags,
'metMin': 0.,
'metSigMin': 0.,
'dPhiJetMet': 0.
}))['cut']
selection[dataOrMC] += "&&" + bJetSelectionL + ">=" + str(
self.nLooseBTags[0])
selection[dataOrMC] += "&&" + zMassSelection
logger.info("Selection " + dataOrMC + ": " +
selection[dataOrMC])
# Calculate yields (take together channels together)
channels = ['MuMu', 'EMu', 'EE']
yield_ttZ_3l = sum(
self.yieldFromCache(
setup, 'TTZ', c, "&&".join(
[lllSelection[c], selection["MC"]]), weight) *
setup.dataLumi[channel] / 1000
for c in ['MuMu', 'EMu', 'EE'])
yield_other = sum(
self.yieldFromCache(
setup, s, c, "&&".join(
[lllSelection[c], selection["MC"]]), weight) *
setup.dataLumi[channel] / 1000
for c in ['MuMu', 'EMu', 'EE']
for s in ['TTJets', 'DY', 'other'])
yield_data_3l = sum(
self.yieldFromCache(
setup, 'Data', c, "&&".join(
[lllSelection[c], selection["Data"]]), "(1)")
for c in ['MuMu', 'EMu', 'EE'])
if not yield_ttZ_3l > 0:
logger.warn("No yield for 3l selection")
estimate = u_float(0, 0)
yield_ttZ_data = yield_data_3l - yield_other
if yield_ttZ_data < 0:
logger.warn("Data-driven ratio is negative!")
yield_ttZ_data = u_float(0, 0)
logger.info("Control region predictions: ")
logger.info(" data: " + str(yield_data_3l))
logger.info(" MC other: " + str(yield_other))
logger.info(" TTZ (MC): " + str(yield_ttZ_3l))
logger.info(" TTZ (data): " + str(yield_ttZ_data))
logger.info(" TTZ (ratio): " +
str(yield_ttZ_data / yield_ttZ_3l))
estimate = (yield_ttZ_data / yield_ttZ_3l) * yield_ttZ_2l
logger.info(" --> " + str(estimate))
return estimate
class SystematicEstimator:
__metaclass__ = abc.ABCMeta
def __init__(self, name, cacheDir=None):
self.name = name
self.initCache(cacheDir)
def initCache(self, cacheDir):
if cacheDir:
self.cacheDir=cacheDir
cacheFileName = os.path.join(cacheDir, self.name+'.pkl')
if not os.path.exists(os.path.dirname(cacheFileName)):
os.makedirs(os.path.dirname(cacheFileName))
self.cache = Cache(cacheFileName, verbosity=2)
else:
self.cache=None
def uniqueKey(self, region, channel, setup):
return region, channel, json.dumps(setup.sys, sort_keys=True), json.dumps(setup.parameters, sort_keys=True), json.dumps(setup.lumi, sort_keys=True)
def cachedEstimate(self, region, channel, setup, save=True):
key = self.uniqueKey(region, channel, setup)
if self.cache and self.cache.contains(key):
res = self.cache.get(key)
logger.debug( "Loading cached %s result for %r : %r"%(self.name, key, res) )
return res
elif self.cache:
return self.cache.add( key, self._estimate( region, channel, setup), save=save)
else:
return self._estimate( region, channel, setup)
@abc.abstractmethod
def _estimate(self, region, channel, setup):
'''Estimate yield in 'region' using setup'''
return
def PUSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup)
up = self.cachedEstimate(region, channel, setup.sysClone({'reweight':['reweightPUUp']}))
down = self.cachedEstimate(region, channel, setup.sysClone({'reweight':['reweightPUDown']}))
return 0.5*(up-down)/ref
def topPtSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup)
up = self.cachedEstimate(region, channel, setup.sysClone({'reweight':['reweightTopPt']}))
return 0.5*(up-ref)/ref
def JERSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup)
up = self.cachedEstimate(region, channel, setup.sysClone({'selectionModifier':'JERUp'}))
down = self.cachedEstimate(region, channel, setup.sysClone({'selectionModifier':'JERDown'}))
return 0.5*(up-down)/ref
def JECSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup)
up = self.cachedEstimate(region, channel, setup.sysClone({'selectionModifier':'JECUp'}))
down = self.cachedEstimate(region, channel, setup.sysClone({'selectionModifier':'JECDown'}))
return 0.5*(up-down)/ref
def leptonFSSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup.sysClone({'reweight':['reweightLeptonFastSimSF']}))
up = self.cachedEstimate(region, channel, setup.sysClone({'reweight':['reweightLeptonFastSimSFUp']}))
down = self.cachedEstimate(region, channel, setup.sysClone({'reweight':['reweightLeptonFastSimSFDown']}))
return 0.5*(up-down)/ref
def btaggingSFbSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup.sysClone({'reweight':['reweightBTag_SF']}))
up = self.cachedEstimate(region, channel, setup.sysClone({'reweight':['reweightBTag_SF_b_Up']}))
down = self.cachedEstimate(region, channel, setup.sysClone({'reweight':['reweightBTag_SF_b_Down']}))
return 0.5*(up-down)/ref
def btaggingSFlSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup.sysClone({'reweight':['reweightBTag_SF']}))
up = self.cachedEstimate(region, channel, setup.sysClone({'reweight':['reweightBTag_SF_l_Up']}))
down = self.cachedEstimate(region, channel, setup.sysClone({'reweight':['reweightBTag_SF_l_Down']}))
return 0.5*(up-down)/ref
def btaggingSFFSSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup.sysClone({'reweight':['reweightBTag_SF']}))
up = self.cachedEstimate(region, channel, setup.sysClone({'reweight':['reweightBTag_SF_FS_Up']}))
down = self.cachedEstimate(region, channel, setup.sysClone({'reweight':['reweightBTag_SF_FS_Down']}))
return 0.5*(up-down)/ref
def getBkgSysJobs(self, region, channel, setup):
l = [
(region, channel, setup.sysClone({'reweight':['reweightPUUp']})),
(region, channel, setup.sysClone({'reweight':['reweightPUDown']})),
(region, channel, setup.sysClone({'reweight':['reweightTopPt']})),
(region, channel, setup.sysClone({'selectionModifier':'JERUp'})),
(region, channel, setup.sysClone({'selectionModifier':'JERDown'})),
(region, channel, setup.sysClone({'selectionModifier':'JECUp'})),
(region, channel, setup.sysClone({'selectionModifier':'JECDown'})),
(region, channel, setup.sysClone({'reweight':['reweightBTag_SF']})),
(region, channel, setup.sysClone({'reweight':['reweightBTag_SF_b_Up']})),
(region, channel, setup.sysClone({'reweight':['reweightBTag_SF_b_Down']})),
(region, channel, setup.sysClone({'reweight':['reweightBTag_SF_l_Up']})),
(region, channel, setup.sysClone({'reweight':['reweightBTag_SF_l_Down']})),
]
return l
def getSigSysJobs(self, region, channel, setup, isFastSim = False):
l = self.getBkgSysJobs(region = region, channel = channel, setup = setup)
if isFastSim:
l.extend( [\
(region, channel, setup.sysClone({'reweight':['reweightBTag_SF_FS_Up']})),
(region, channel, setup.sysClone({'reweight':['reweightBTag_SF_FS_Down']})),
(region, channel, setup.sysClone({'reweight':['reweightLeptonFastSimSF']})),
(region, channel, setup.sysClone({'reweight':['reweightLeptonFastSimSFUp']})),
(region, channel, setup.sysClone({'reweight':['reweightLeptonFastSimSFDown']})),
] )
return l
class SystematicBaseClass:
__metaclass__ = abc.ABCMeta
def __init__(self, name, cacheDir=None):
self.name = name
self.initCache(cacheDir)
def initCache(self, cacheDir):
if cacheDir:
self.cacheDir = cacheDir
cacheFileName = os.path.join(cacheDir, self.name + '.pkl')
if not os.path.exists(os.path.dirname(cacheFileName)):
os.makedirs(os.path.dirname(cacheFileName))
self.cache = Cache(cacheFileName, verbosity=2)
else:
self.cache = None
def uniqueKey(self, region, channel, setup):
return region, channel, json.dumps(
setup.sys, sort_keys=True), json.dumps(setup.parameters,
sort_keys=True), json.dumps(
setup.lumi,
sort_keys=True)
def cachedEstimate(self, region, channel, setup, save=True):
key = self.uniqueKey(region, channel, setup)
if self.cache and self.cache.contains(key):
res = self.cache.get(key)
if setup.verbose:
print "Loading cached %s result for %r : %r" % (self.name, key,
res)
return res
elif self.cache:
return self.cache.add(key,
self._estimate(region, channel, setup),
save=save)
else:
return self._estimate(region, channel, setup)
@abc.abstractmethod
def _estimate(self, region, channel, setup):
'''Estimate yield in 'region' using setup'''
return
def PUSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup)
up = self.cachedEstimate(region, channel,
setup.sysClone({'weight': 'weightPUUp'}))
down = self.cachedEstimate(region, channel,
setup.sysClone({'weight': 'weightPUDown'}))
return 0.5 * (up - down) / ref
def topPtSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup)
up = self.cachedEstimate(
region, channel, setup.sysClone({'reweight': ['reweightTopPt']}))
return 0.5 * (up - ref) / ref
def JERSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup)
up = self.cachedEstimate(
region, channel, setup.sysClone({'selectionModifier': 'JERUp'}))
down = self.cachedEstimate(
region, channel, setup.sysClone({'selectionModifier': 'JERDown'}))
return 0.5 * (up - down) / ref
def JECSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup)
up = self.cachedEstimate(
region, channel, setup.sysClone({'selectionModifier': 'JECUp'}))
down = self.cachedEstimate(
region, channel, setup.sysClone({'selectionModifier': 'JECDown'}))
return 0.5 * (up - down) / ref
def leptonFSSystematic(self, region, channel, setup):
ref = self.cachedEstimate(
region, channel,
setup.sysClone({'reweight': ['reweightLeptonFastSimSF']}))
up = self.cachedEstimate(
region, channel,
setup.sysClone({'reweight': ['reweightLeptonFastSimSFUp']}))
down = self.cachedEstimate(
region, channel,
setup.sysClone({'reweight': ['reweightLeptonFastSimSFDown']}))
return 0.5 * (up - down) / ref
def btaggingSFbSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel,
setup.sysClone({'useBTagWeights': 'SF'}))
up = self.cachedEstimate(region, channel,
setup.sysClone({'useBTagWeights': 'SF_b_Up'}))
down = self.cachedEstimate(
region, channel, setup.sysClone({'useBTagWeights': 'SF_b_Down'}))
return 0.5 * (up - down) / ref
def btaggingSFlSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel,
setup.sysClone({'useBTagWeights': 'SF'}))
up = self.cachedEstimate(region, channel,
setup.sysClone({'useBTagWeights': 'SF_l_Up'}))
down = self.cachedEstimate(
region, channel, setup.sysClone({'useBTagWeights': 'SF_l_Down'}))
return 0.5 * (up - down) / ref
def btaggingSFFSSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel,
setup.sysClone({'useBTagWeights': 'SF'}))
up = self.cachedEstimate(
region, channel, setup.sysClone({'useBTagWeights': 'SF_FS_Up'}))
down = self.cachedEstimate(
region, channel, setup.sysClone({'useBTagWeights': 'SF_FS_Down'}))
return 0.5 * (up - down) / ref
def getBkgSysJobs(self, region, channel, setup):
return [
(region, channel, setup.sysClone({'weight': 'weightPUUp'})),
(region, channel, setup.sysClone({'weight': 'weightPUDown'})),
(region, channel, setup.sysClone({'reweight': ['reweightTopPt']})),
(region, channel, setup.sysClone({'selectionModifier': 'JERUp'})),
(region, channel, setup.sysClone({'selectionModifier':
'JERDown'})),
(region, channel, setup.sysClone({'selectionModifier': 'JECUp'})),
(region, channel, setup.sysClone({'selectionModifier':
'JECDown'})),
# (region, channel, setup.sysClone({'reweight':['reweightLeptonFastSimSF']})),
# (region, channel, setup.sysClone({'reweight':['reweightLeptonFastSimSFUp']})),
# (region, channel, setup.sysClone({'reweight':['reweightLeptonFastSimSFDown']})),
(region, channel, setup.sysClone({'useBTagWeights': 'SF'})),
(region, channel, setup.sysClone({'useBTagWeights': 'SF_b_Up'})),
(region, channel, setup.sysClone({'useBTagWeights': 'SF_b_Down'})),
(region, channel, setup.sysClone({'useBTagWeights': 'SF'})),
(region, channel, setup.sysClone({'useBTagWeights': 'SF_l_Up'})),
(region, channel, setup.sysClone({'useBTagWeights': 'SF_l_Down'})),
(region, channel, setup.sysClone({'useBTagWeights': 'SF'})),
# (region, channel, setup.sysClone({'useBTagWeights':'SF_FS_Up'})),
# (region, channel, setup.sysClone({'useBTagWeights':'SF_FS_Down'})),
]
def getSigSysJobs(self, region, channel, setup):
return [
(region, channel,
setup.sysClone({'reweight': ['reweightLeptonFastSimSF']})),
(region, channel,
setup.sysClone({'reweight': ['reweightLeptonFastSimSFUp']})),
(region, channel,
setup.sysClone({'reweight': ['reweightLeptonFastSimSFDown']})),
(region, channel, setup.sysClone({'useBTagWeights': 'SF_FS_Up'})),
(region, channel, setup.sysClone({'useBTagWeights':
'SF_FS_Down'})),
]
class SystematicBaseClass:
__metaclass__ = abc.ABCMeta
def __init__(self, name, cacheDir=None):
self.name = name
self.initCache(cacheDir)
def initCache(self, cacheDir):
if cacheDir:
self.cacheDir = cacheDir
cacheFileName = os.path.join(cacheDir, self.name + ".pkl")
if not os.path.exists(os.path.dirname(cacheFileName)):
os.makedirs(os.path.dirname(cacheFileName))
self.cache = Cache(cacheFileName, verbosity=2)
else:
self.cache = None
def uniqueKey(self, region, channel, setup):
return (
region,
channel,
json.dumps(setup.sys, sort_keys=True),
json.dumps(setup.parameters, sort_keys=True),
json.dumps(setup.lumi, sort_keys=True),
)
def cachedEstimate(self, region, channel, setup, save=True):
key = self.uniqueKey(region, channel, setup)
if self.cache and self.cache.contains(key):
res = self.cache.get(key)
if setup.verbose:
print "Loading cached %s result for %r : %r" % (self.name, key, res)
return res
elif self.cache:
return self.cache.add(key, self._estimate(region, channel, setup), save=save)
else:
return self._estimate(region, channel, setup)
@abc.abstractmethod
def _estimate(self, region, channel, setup):
"""Estimate yield in 'region' using setup"""
return
def PUSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup)
up = self.cachedEstimate(region, channel, setup.sysClone({"weight": "weightPUUp"}))
down = self.cachedEstimate(region, channel, setup.sysClone({"weight": "weightPUDown"}))
return 0.5 * (up - down) / ref
def topPtSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup)
up = self.cachedEstimate(region, channel, setup.sysClone({"reweight": ["reweightTopPt"]}))
return 0.5 * (up - ref) / ref
def JERSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup)
up = self.cachedEstimate(region, channel, setup.sysClone({"selectionModifier": "JERUp"}))
down = self.cachedEstimate(region, channel, setup.sysClone({"selectionModifier": "JERDown"}))
return 0.5 * (up - down) / ref
def JECSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup)
up = self.cachedEstimate(region, channel, setup.sysClone({"selectionModifier": "JECUp"}))
down = self.cachedEstimate(region, channel, setup.sysClone({"selectionModifier": "JECDown"}))
return 0.5 * (up - down) / ref
def leptonFSSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup.sysClone({"reweight": ["reweightLeptonFastSimSF"]}))
up = self.cachedEstimate(region, channel, setup.sysClone({"reweight": ["reweightLeptonFastSimSFUp"]}))
down = self.cachedEstimate(region, channel, setup.sysClone({"reweight": ["reweightLeptonFastSimSFDown"]}))
return 0.5 * (up - down) / ref
def btaggingSFbSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup.sysClone({"useBTagWeights": "SF"}))
up = self.cachedEstimate(region, channel, setup.sysClone({"useBTagWeights": "SF_b_Up"}))
down = self.cachedEstimate(region, channel, setup.sysClone({"useBTagWeights": "SF_b_Down"}))
return 0.5 * (up - down) / ref
def btaggingSFlSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup.sysClone({"useBTagWeights": "SF"}))
up = self.cachedEstimate(region, channel, setup.sysClone({"useBTagWeights": "SF_l_Up"}))
down = self.cachedEstimate(region, channel, setup.sysClone({"useBTagWeights": "SF_l_Down"}))
return 0.5 * (up - down) / ref
def btaggingSFFSSystematic(self, region, channel, setup):
ref = self.cachedEstimate(region, channel, setup.sysClone({"useBTagWeights": "SF"}))
up = self.cachedEstimate(region, channel, setup.sysClone({"useBTagWeights": "SF_FS_Up"}))
down = self.cachedEstimate(region, channel, setup.sysClone({"useBTagWeights": "SF_FS_Down"}))
return 0.5 * (up - down) / ref
def getBkgSysJobs(self, region, channel, setup):
return [
(region, channel, setup.sysClone({"weight": "weightPUUp"})),
(region, channel, setup.sysClone({"weight": "weightPUDown"})),
(region, channel, setup.sysClone({"reweight": ["reweightTopPt"]})),
(region, channel, setup.sysClone({"selectionModifier": "JERUp"})),
(region, channel, setup.sysClone({"selectionModifier": "JERDown"})),
(region, channel, setup.sysClone({"selectionModifier": "JECUp"})),
(region, channel, setup.sysClone({"selectionModifier": "JECDown"})),
# (region, channel, setup.sysClone({'reweight':['reweightLeptonFastSimSF']})),
# (region, channel, setup.sysClone({'reweight':['reweightLeptonFastSimSFUp']})),
# (region, channel, setup.sysClone({'reweight':['reweightLeptonFastSimSFDown']})),
(region, channel, setup.sysClone({"useBTagWeights": "SF"})),
(region, channel, setup.sysClone({"useBTagWeights": "SF_b_Up"})),
(region, channel, setup.sysClone({"useBTagWeights": "SF_b_Down"})),
(region, channel, setup.sysClone({"useBTagWeights": "SF"})),
(region, channel, setup.sysClone({"useBTagWeights": "SF_l_Up"})),
(region, channel, setup.sysClone({"useBTagWeights": "SF_l_Down"})),
(region, channel, setup.sysClone({"useBTagWeights": "SF"})),
# (region, channel, setup.sysClone({'useBTagWeights':'SF_FS_Up'})),
# (region, channel, setup.sysClone({'useBTagWeights':'SF_FS_Down'})),
]
def getSigSysJobs(self, region, channel, setup):
return [
(region, channel, setup.sysClone({"reweight": ["reweightLeptonFastSimSF"]})),
(region, channel, setup.sysClone({"reweight": ["reweightLeptonFastSimSFUp"]})),
(region, channel, setup.sysClone({"reweight": ["reweightLeptonFastSimSFDown"]})),
(region, channel, setup.sysClone({"useBTagWeights": "SF_FS_Up"})),
(region, channel, setup.sysClone({"useBTagWeights": "SF_FS_Down"})),
]
class DataDrivenTTZEstimate(SystematicEstimator):
def __init__(self, name, cacheDir=None, useTop16009=False):
super(DataDrivenTTZEstimate, self).__init__(name, cacheDir=cacheDir)
self.nJets = (3,-1) # jet selection (min, max)
self.nLooseBTags = (2,-1) # loose bjet selection (min, max)
self.nMediumBTags = (0,-1) # bjet selection (min, max)
self.useTop16009 = useTop16009
self.ratioTop16009 = 1.27 #
self.sysErrTop16009 = (-0.17, +0.20)
self.statErrTop16009 = (-0.37, +0.42)
# Because we are going to reuse a lot of yields which otherwise will be terribly slow
self.helperCacheName = os.path.join('.', 'helperCache.pkl')
self.helperCache = Cache(self.helperCacheName, verbosity=2)
def yieldFromCache(self, setup, sample, channel, selectionString, weightString):
s = (sample, channel, selectionString, weightString)
if self.helperCache.contains(s):
return self.helperCache.get(s)
else:
yieldFromDraw = u_float(**setup.sample[sample][channel].getYieldFromDraw(selectionString, weightString))
self.helperCache.add(s, yieldFromDraw, save=True)
return yieldFromDraw
#Concrete implementation of abstract method 'estimate' as defined in Systematic
def _estimate(self, region, channel, setup):
logger.info("Data-driven TTZ estimate for region " + str(region) + " in channel " + channel + " and setup " + str(setup.sys) + ":")
#Sum of all channels for 'all'
if channel=='all':
estimate = sum([self.cachedEstimate(region, c, setup) for c in ['MuMu', 'EE', 'EMu']])
else:
zWindow= 'allZ' if channel=='EMu' else 'offZ'
preSelection = setup.preselection('MC', zWindow=zWindow, channel=channel)
MC_2l = "&&".join([region.cutString(setup.sys['selectionModifier']), preSelection['cut']])
weight = setup.weightString()
logger.info("weight: %s", weight)
yield_ttZ_2l = setup.lumi[channel]/1000.*self.yieldFromCache(setup, 'TTZ', channel, MC_2l, weight)
logger.info("yield_MC_2l: %s"%yield_ttZ_2l)
if self.useTop16009:
sysError = max((abs(x) for x in self.sysErrTop16009)) # not sure yet to handle assymetric errors
statError = max((abs(x) for x in self.statErrTop16009))
error = sqrt(sysError*sysError+statError*statError)
return u_float(self.ratioTop16009, error)*yield_ttZ_2l
else:
# pt leptons > 30, 20, 10 GeV
useTrigger = False # setup.parameters['useTriggers'] # better not to use three lepton triggers, seems to be too inefficient
lllSelection = {}
lllSelection['MuMu'] = "&&".join([getLeptonString(3, 0), getPtThresholdString(30, 20, 10)]) + ("&&HLT_3mu" if useTrigger else "")
lllSelection['MuMuE'] = "&&".join([getLeptonString(2, 1), getPtThresholdString(30, 20, 10)]) + ("&&HLT_2mu1e" if useTrigger else "")
lllSelection['MuEE'] = "&&".join([getLeptonString(1, 2), getPtThresholdString(30, 20, 10)]) + ("&&HLT_2e1mu" if useTrigger else "")
lllSelection['EE'] = "&&".join([getLeptonString(0, 3), getPtThresholdString(30, 20, 10)]) + ("&&HLT_3e" if useTrigger else "")
lllSelection['EMu'] = "(("+lllSelection['MuMuE']+")||("+lllSelection['MuEE']+"))"
bJetSelectionM = "(Sum$(JetGood_pt>30&&abs(JetGood_eta)<2.4&&JetGood_id&&JetGood_btagCSV>0.890))"
bJetSelectionL = "(Sum$(JetGood_pt>30&&abs(JetGood_eta)<2.4&&JetGood_id&&JetGood_btagCSV>0.605))"
zMassSelection = "abs(mlmZ_mass-91.1876)<10"
# Start from base hadronic selection and add loose b-tag and Z-mass requirement
selection = {}
for dataOrMC in ["Data", "MC"]:
selection[dataOrMC] = setup.selection(dataOrMC, hadronicSelection = True, **setup.defaultParameters(update={'nJets': self.nJets, 'nBTags':self.nMediumBTags, 'metMin': 0., 'metSigMin':0., 'dPhiJetMet':0. }))['cut']
selection[dataOrMC] += "&&" + bJetSelectionL+">="+str(self.nLooseBTags[0])
selection[dataOrMC] += "&&" + zMassSelection
logger.info("Selection " + dataOrMC + ": " + selection[dataOrMC])
# Calculate yields (take together channels together)
channels = ['MuMu','EMu','EE']
yield_ttZ_3l = sum(self.yieldFromCache(setup, 'TTZ', c, "&&".join([lllSelection[c], selection["MC"]]), weight)*setup.dataLumi[channel]/1000 for c in ['MuMu','EMu','EE'])
yield_other = sum(self.yieldFromCache(setup, s, c, "&&".join([lllSelection[c], selection["MC"]]), weight)*setup.dataLumi[channel]/1000 for c in ['MuMu','EMu','EE'] for s in ['TTJets', 'DY', 'other'])
yield_data_3l = sum(self.yieldFromCache(setup, 'Data', c, "&&".join([lllSelection[c], selection["Data"]]), "(1)") for c in ['MuMu','EMu','EE'])
if not yield_ttZ_3l > 0:
logger.warn("No yield for 3l selection")
estimate = u_float(0, 0)
yield_ttZ_data = yield_data_3l - yield_other
if yield_ttZ_data < 0:
logger.warn("Data-driven ratio is negative!")
yield_ttZ_data = u_float(0, 0)
logger.info("Control region predictions: ")
logger.info(" data: " + str(yield_data_3l))
logger.info(" MC other: " + str(yield_other))
logger.info(" TTZ (MC): " + str(yield_ttZ_3l))
logger.info(" TTZ (data): " + str(yield_ttZ_data))
logger.info(" TTZ (ratio): " + str(yield_ttZ_data/yield_ttZ_3l))
estimate = (yield_ttZ_data/yield_ttZ_3l)*yield_ttZ_2l
logger.info(" --> " + str(estimate))
return estimate
