def hist2D(self, var1, nBins1, a1, b1, var2, nBins2, a2, b2, **kwargs):
name = kwargs.get(
'name',
makeHistName(self.label + "_merged", "%s_vs_%s" % (var1, var2)))
title = kwargs.get('title', self.label)
blind = kwargs.get('blind', self.blind)
kwargs['scale'] = self.scale * kwargs.get('scale',
1.0) # pass scale down
verbosity = kwargs.get('verbosity', 0)
printVerbose(
">>>\n>>> Samples - %s, %s vs. %s: %s" %
(color(name, color="grey"), var1, var2, self.filenameshort),
verbosity)
printVerbose(">>> scale: %.4f" % (kwargs['scale']), verbosity)
hist2D = TH2D(name, title, nBins2, a2, b2, nBins1, a1, b1)
for sample in self.samples:
if 'name' in kwargs: # prevent memory leaks
kwargs['name'] = makeHistName(
sample.label, name.replace(self.label + '_', ''))
hist2D.Add(
sample.hist2D(var1, nBins1, a1, b1, var2, nBins2, a2, b2,
**kwargs))
return hist2D
def normalizeSignal(self,S_exp,**kwargs):
"""Calculates normalization for a given expected signal yield."""
verbosity = kwargs.get('verbosity',0)
cuts = [ ("%s && %s" % (baseline, category1)),
("%s && %s" % (baseline, category2)), ]
cuts = kwargs.get('cuts',cuts)
weight = kwargs.get('weight',"")
(aa,bb) = kwargs.get('signalregion',(0,40))
N = 0; MC = 0
scale = 1
for i,cut in enumerate(cuts):
cut = combineCuts("m_sv>0", cut)
name = "m_sv_for_signal_renormalization_%d" % i
hist = self.hist("m_sv",100,aa,bb,name=name,cuts=cut,weight=weight)
N += hist.GetSumOfWeights()
MC += hist.GetEntries()
gDirectory.Delete(name)
if N:
scale = S_exp / N * self.scale
printVerbose(">>> normalizeSignal: S_exp=%.4f, N=%.4f, MC=%.1f, old scale=%.4f, scale=%.4f" % (S_exp, N, MC, self.scale, scale), verbosity)
printVerbose(">>> normalizeSignal: signalregion=(%.1f,%.1f)" % (aa,bb),verbosity)
else: print warning("Could not find normalization for signal: no MC events in given signal region after cuts (%s)!" % cuts)
self.setAllScales(scale)
return scale
def normalizeSignal(self, S_exp, **kwargs):
"""Calculates normalization for a given expected signal yield."""
if not self.isSignal:
print warning("normalizeSignal: Not a signal sample!")
verbosity = kwargs.get('verbosity', 0)
var = kwargs.get('var', "m_sv")
cuts = [
("%s && %s" % (baseline, category1)),
("%s && %s" % (baseline, category2)),
]
cuts = kwargs.get('cuts', cuts)
if not isinstance(cuts, list) and not isinstance(cuts, tuple):
cuts = [cuts]
(aa, bb) = kwargs.get('signalregion', (0, 40))
weight = kwargs.get('weight', "")
channel = kwargs.get('channel', "mutau")
#treeName = kwargs.get('treeName',"tree_%s"%channel)
setScale = kwargs.get('setScale', True)
N = 0
MC = 0
scale = 1
for i, cut in enumerate(cuts):
#cut = combineCuts("%s<%s && %s<%s"%(aa,var,var,bb), cut) # remove over and underflow
name = "m_sv_for_signal_renormalization_%d" % i
hist = self.hist("m_sv",
100,
aa,
bb,
name=name,
cuts=cut,
weight=weight,
verbosity=verbosity)
N += hist.GetSumOfWeights()
MC += hist.GetEntries()
gDirectory.Delete(name)
printVerbose(">>> normalizeSignal: N=%s, MC=%s" % (N, MC),
verbosity)
if N:
scale = S_exp / N * self.scale
printVerbose(
">>> normalizeSignal: S_exp=%.4f, N=%.4f, MC=%.1f, old scale=%.4f, scale=%.4f"
% (S_exp, N, MC, self.scale, scale), verbosity)
#printVerbose(">>> normalizeSignal: signalregion=(%.1f,%.1f)" % (aa,bb),verbosity)
else:
print warning(
"Could not find normalization for signal: no MC events (N=%s,MC=%s) in given signal region after cuts (%s)!"
% (N, MC, cuts))
if setScale: self.setAllScales(scale)
return scale
def hist2D(self, var1, nBins1, a1, b1, var2, nBins2, a2, b2, **kwargs):
"""Make a 2D histogram with a tree."""
scale = kwargs.get('scale', 1.0) * self.scale
tree = self.file.Get(self.treeName)
name = kwargs.get('name', makeHistName(self.label, "%s_vs_%s" % (var1,var2)))
title = kwargs.get('title', self.label)
verbosity = kwargs.get('verbosity', 0)
blindcuts = ""
if var1 in self.blind: blindcuts += self.blind[var1]
if var2 in self.blind: blindcuts += self.blind[var2]
weight = combineWeights(self.weight, kwargs.get('weight', ""))
cuts = combineCuts(self.cuts, kwargs.get('cuts', ""), blindcuts, weight=weight)
printVerbose(">>>>\n>>> Sample - %s, %s vs. %s: %s" % (color(name,color="grey"), var1, var2, self.filenameshort),verbosity)
printVerbose(">>> scale: %.4f" % (scale),verbosity)
printVerbose(">>> weight: %s" % (weight),verbosity)
printVerbose(">>> %s" % (cuts),verbosity)
hist2D = TH2F(name, title, nBins2, a2, b2, nBins1, a1, b1)
out = tree.Draw("%s:%s >> %s" % (var1,var2,name), cuts, "gOff")
if out < 0: print error("Drawing histogram for %s sample failed!" % (title))
#if scale is not 1.0: hist.Scale(scale)
#if scale is 0.0: print warning("Scale of %s is 0!" % self.label)
return hist2D
def calculateLumiAcceptance(self,cuts,**kwargs):
"""Calculates scale for a given expected signal yield, to divide
out the luminosity and acceptance. This method only returns the scale,
it does not rescale the signal."""
verbosity = kwargs.get('verbosity',0)
weight = kwargs.get('weight',"")
(a,b) = kwargs.get('range',(0,500))
scale = 1
N_tot = self.N
name = "m_sv_for_LumiAcceptance"
hist = self.hist("m_sv",100,a,b,name=name,cuts=cuts,weight=weight)
(N,MC) = (hist.GetSumOfWeights(),hist.GetEntries())
gDirectory.Delete(name)
if N_tot and N and lumi:
scale = N_tot/(N*lumi)
printVerbose(">>> normalizeSignal: N_tot=%.4f, N=%.4f, MC=%.1f, lumi=%s, current scale=%.4f, scale=%.4f" % (N_tot, N, MC, lumi, self.scale, scale), verbosity)
printVerbose(">>> normalizeSignal: range=(%.1f,%.1f)" % (a,b),verbosity)
else: print warning("Could not find normalization for signal: N_tot=%s, N=%s, lumi=%s!" % (N_tot,N,lumi))
return scale
def hist(self, var, nBins, a, b, **kwargs):
name = kwargs.get('name', makeHistName(self.label + "_merged", var))
name += kwargs.get('append_name', "")
title = kwargs.get('title', self.label)
blind = kwargs.get('blind', self.blind)
kwargs['scale'] = self.scale * kwargs.get('scale',
1.0) # pass scale down
verbosity = kwargs.get('verbosity', 0)
printVerbose(
">>>\n>>> Samples - %s, %s: %s" %
(color(name, color="grey"), var, self.filenameshort), verbosity)
printVerbose(">>> scale: %.4f" % (kwargs['scale']), verbosity)
hist = TH1D(name, title, nBins, a, b)
hist.Sumw2()
for sample in self.samples:
if 'name' in kwargs: # prevent memory leaks
kwargs['name'] = makeHistName(
sample.label, name.replace(self.label + '_', ''))
hist_new = sample.hist(var, nBins, a, b, **kwargs)
hist.Add(hist_new)
printVerbose(
">>> sample %s added with %.1f events (%d entries)" %
(sample.label, hist_new.Integral(), hist_new.GetEntries()),
verbosity,
level=2)
if verbosity > 2: printBinError(hist)
return hist
def calculateLumiAcceptance(self, cuts, **kwargs):
"""Calculates scale for a given expected signal yield, to divide
out the luminosity and acceptance. This method only returns the scale,
it does not rescale the signal."""
verbosity = kwargs.get('verbosity', 0)
var = kwargs.get('var', "m_sv")
weight = kwargs.get('weight', "")
(a, b) = kwargs.get('signalregion', (0, 500))
scale = 1
N_tot = self.N
name = "%s_for_LA" % var
hist = self.hist(var, 100, a, b, name=name, cuts=cuts, weight=weight)
(N, MC) = (hist.GetSumOfWeights(), hist.GetEntries())
gDirectory.Delete(name)
#cuts = combineCuts("%s<%s && %s<%s"%(a,var,var,b), cuts)
printVerbose(">>> calculateLA:", verbosity)
printVerbose(">>> cuts=%s" % (cuts), verbosity)
if N_tot and N and lumi:
scale = N_tot / (N * lumi * 1000)
printVerbose(
">>> N_tot=%.4f, N=%.4f, MC=%.1f, lumi=%s, current scale=%.4f, scale=%.4f"
% (N_tot, N, MC, lumi, self.scale, scale), verbosity)
#printVerbose(">>> signalregion=(%.1f,%.1f)" % (a,b),verbosity)
else:
print warning(
"Could not find normalization for signal: N_tot=%s, N=%s, lumi=%s!"
% (N_tot, N, lumi))
return scale
def histAndColor(self, var, nBins, a, b, **kwargs):
'''Return a list of tuples containing a histogram and a color.
Return multiple ntuples if a sample need to be split.'''
split = kwargs.get('split', False) and len(self.split)
verbosity = kwargs.get('verbosity', 0)
if split:
printVerbose(">>> histAndColor: splitting %s" % (self.label),
verbosity)
histsAndColors = []
cuts0 = kwargs.get('cuts', "")
for key, (splitlabel, splitcut,
splitcolor) in self.split.iteritems():
kwargs['cuts'] = combineCuts(cuts0, splitcut)
kwargs['title'] = splitlabel
kwargs['append_name'] = "_%s" % (key)
hist = self.hist(var, nBins, a, b, **kwargs)
histsAndColors.append((hist, splitcolor))
return histsAndColors
else:
printVerbose(">>> histAndColor: not splitting", verbosity, level=2)
hist = self.hist(var, nBins, a, b, **kwargs)
return [(hist, self.color)]
def renormalizeWJ(samples, **kwargs):
"""Helpfunction to renormalize W + Jets."""
#print ">>>\n>>> renormalizing WJ"
var = kwargs.get('var', "pfmt_1")
QCD = kwargs.get('QCD', True)
channel = kwargs.get('channel', "mutau")
label = kwargs.get('label', "baseline")
cuts = kwargs.get('cuts', baseline)
reset = kwargs.get('reset', True)
shift_QCD = kwargs.get('shift_QCD', 0) # e.g. 0.30
prepend = kwargs.get('prepend', "")
verbosity = kwargs.get('verbosity', 0)
ratio_WJ_QCD_SS = kwargs.get('ratio_WJ_QCD_SS', True)
#samples = kwargs.get('samples',[ ])
name = "%s/%s%s/%s_tail_%s_noWJrenormalization.png" % (PLOTS_DIR, channel,
mylabel, var, label)
title = "%s: %s" % (channel.replace("tau", "#tau").replace("mu",
"#mu"), label)
printVerbose(">>> WJ renormalization with:\n>>> %s: %s\n>>> %s: %s" %
("QCD", QCD, "ratio_WJ_QCD_SS", ratio_WJ_QCD_SS),
verbosity,
level=2)
plot = Plot(samples,
var,
200,
80,
200,
cuts=cuts,
QCD=QCD,
ratio_WJ_QCD_SS=ratio_WJ_QCD_SS,
reset=True,
shift_QCD=shift_QCD,
verbosity=verbosity)
#plot.plot(stack=True, title=title, staterror=True, ratio=True)
scale = plot.renormalizeWJ(prepend=prepend, verbosity=verbosity)
plot.close()
def hist(self, var, nBins, a, b, **kwargs):
"""Make a histogram with a tree."""
scale = kwargs.get('scale', 1.0) * self.scale
tree = self.file.Get(self.treeName)
name = kwargs.get('name', makeHistName(self.label, var))
title = kwargs.get('title', self.label)
shift = kwargs.get('shift', 0)
smear = kwargs.get('smear', 0)
verbosity = kwargs.get('verbosity', 0)
if self.isSignal and self.scale is not self.scaleBU and self.scaleBU:
title += " (#times%d)" % (self.scale/self.scaleBU)
blindcuts = ""
if var in self.blind: blindcuts = self.blind[var]
weight = combineWeights(self.weight, kwargs.get('weight', ""))
cuts = combineCuts(self.cuts, kwargs.get('cuts', ""), blindcuts, weight=weight)
hist = TH1F(name, title, nBins, a, b)
out = tree.Draw("%s >> %s" % (var,name), cuts, "gOff")
if shift or smear:
mean0 = hist.GetMean()
#smear = min(1,smear)
#smear = sqrt(smear*smear-1) #*sigma
#tree.SetAlias("rng","sin(2*pi*rndm)*sqrt(-2*log(rndm))")
var2 = "%s*%s + %s + %s*%s" % (var,smear,shift,(1-smear),mean0)
tree.Draw("%s >> %s" % (var2,name), cuts, "gOff")
if out < 0: print error("Drawing histogram for %s sample failed!" % (title))
if scale is not 1.0: hist.Scale(scale)
if scale is 0.0: print warning("Scale of %s is 0!" % self.label)
#print hist.GetEntries()
#gDirectory.Delete(label)
printVerbose(">>>\n>>> Sample - %s, %s: %s" % (color(name,color="grey"), var, self.filenameshort),verbosity)
printVerbose(">>> scale: %.4f (%.4f)" % (scale,self.scale),verbosity)
printVerbose(">>> weight: %s" % (("\n>>>%s*("%(' '*18)).join(weight.rsplit('*(',max(0,weight.count("*(")-1)))),verbosity)
printVerbose(">>> %s" % (cuts.replace("*(","\n>>>%s*("%(' '*18))),verbosity)
return hist
def mergeSamples(sample_list, names, **kwargs):
"""Merge samples"""
verbosity = kwargs.get('verbosity', 2)
if not isinstance(names, list): names = [names]
name0 = kwargs.get('name', names[0]) #+ " merged"
signal = kwargs.get('signal', False)
background = kwargs.get('background', True) and not signal
labels = kwargs.get('labels', []) # extra search term
labels.append(kwargs.get('label', ""))
color0 = kwargs.get('color',
colors_dict.get(name0.replace("_merged", ''), kBlack))
samples = Samples(name0,
background=background,
signal=signal,
color=color0)
printVerbose(">>>", verbosity, level=2)
printVerbose(">>> merging %s" % (name0), verbosity, level=1)
# get samples containing names and label
merge_list = []
for name in names:
merge_list += [s for s in sample_list if s.isPartOf(name, *labels)]
# check if sample list of contains to-be-stitched-sample
if len(merge_list) < 2:
print warning("Could not stitch %s: less than two %s samples" %
(name0, name0))
#return sample_list
fill = max([len(s.label) for s in merge_list])
# add samples with name0 and label
for sample in merge_list:
printVerbose(">>> merging %s to %s: %s" %
(sample.label.ljust(fill), name0, sample.filenameshort),
verbosity,
level=2)
samples.add(sample)
# remove merged samples from sample_list
if samples.samples:
sample_list.append(samples)
#print "samples.samples.label = %s\n" % [s.label for s in samples.samples]
#print "sample_list.label = %s\n" % [s.label for s in sample_list]
for sample in samples.samples:
#print "sample.name =", sample.label
sample_list.remove(sample)
def hist(self, var, nBins, a, b, **kwargs):
"""Make a histogram with a tree."""
scale = kwargs.get('scale', 1.0) * self.scale
treeName = kwargs.get('treeName', self.treeName)
name = kwargs.get('name', makeHistName(self.label, var))
name += kwargs.get('append_name', "")
title = kwargs.get('title', self.label)
shift = kwargs.get('shift', 0)
smear = kwargs.get('smear', 0)
blind = kwargs.get('blind', self.blind)
verbosity = kwargs.get('verbosity', 0)
if self.isSignal and self.scale is not self.scaleBU and self.scaleBU:
title += " (#times%d)" % (self.scale / self.scaleBU)
blindcuts = ""
if var in blind and "SS" not in name:
blindcuts = blind[
var] # TODO: blind by removing bins from hist or rounding? FindBin(a), SetBinContent
weight = combineWeights(self.weight, kwargs.get('weight', ""))
cuts = combineCuts(self.cuts,
kwargs.get('cuts', ""),
blindcuts,
weight=weight)
tree = self.file.Get(treeName)
if not tree or not isinstance(tree, TTree):
print error("Could not find tree \"%s\" for %s! Check %s" %
(treeName, self.label, self.filenameshort))
hist = TH1D(name, title, nBins, a, b)
hist.Sumw2()
out = tree.Draw("%s >> %s" % (var, name), cuts, "gOff")
if shift or (smear and smear != 1):
mean0 = hist.GetMean()
#smear = min(1,smear)
#smear = sqrt(smear*smear-1) #*sigma
#tree.SetAlias("rng","sin(2*pi*rndm)*sqrt(-2*log(rndm))")
var2 = "%s*%s + %s + %s*%s" % (var, smear, shift,
(1 - smear), mean0)
tree.Draw("%s >> %s" % (var2, name), cuts, "gOff")
if out < 0:
print error("Drawing histogram for %s sample failed!" % (title))
if scale is not 1.0: hist.Scale(scale)
if scale is 0.0: print warning("Scale of %s is 0!" % self.label)
if verbosity > 2: printBinError(hist)
#print hist.GetEntries()
#gDirectory.Delete(label)
printVerbose(
">>>\n>>> Sample - %s, %s: %s (%s)" % (color(
name, color="grey"), var, self.filenameshort, self.treeName),
verbosity)
printVerbose(">>> scale: %.4f (%.4f)" % (scale, self.scale),
verbosity)
printVerbose(
">>> weight: %s" % (("\n>>>%s*(" % (' ' * 18)).join(
weight.rsplit('*(', max(0,
weight.count("*(") - 1)))), verbosity)
printVerbose(
">>> entries: %d (%.2f integral)" %
(hist.GetEntries(), hist.Integral()), verbosity)
printVerbose(
">>> %s" % (cuts.replace("*(", "\n>>>%s*(" % (' ' * 18))),
verbosity)
return hist
def stitchSamples(sample_list, name0, **kwargs):
"""Stitching samples: merge samples
and reweight inclusive sample and rescale jet-binned samples"""
verbosity = kwargs.get('verbosity', 2)
printVerbose(">>>", verbosity, level=2)
printVerbose(">>> stiching %s: rescale, reweight and merge samples" %
(name0),
verbosity,
level=1)
# see /shome/ytakahas/work/TauTau/SFrameAnalysis/TauTauResonances/plot/config.py
# DY cross sections 5765.4 [ 4954.0, 1012.5, 332.8, 101.8, 54.8 ]
# WJ cross sections 61526.7 [ 50380.0, 9644.5, 3144.5, 954.8, 485.6 ]
sigmasLO = {
"DY": {
"M-50": 4954.0,
"M-10to50": 18610.0
},
"WJ": {
"": 50380.0
}
}
sigmasNLO = {
"DY": {
"M-50": 5765.4,
"M-10to50": 21658.0
},
"WJ": {
"": 61526.7
}
}
#name0 = "DY" #"WJ"
label_incl = kwargs.get('label_incl', "Jets")
name_incl = kwargs.get('name_incl', name0 + label_incl)
labels = kwargs.get('labels', []) # extra search term
labels.append(kwargs.get('label', ""))
sigmaLO = sigmasLO[name0][labels[0]]
kfactor = sigmasNLO[name0][labels[0]] / sigmaLO
N_incl = 0
weights = []
stitch_list = [s for s in sample_list if s.isPartOf(name0, *labels)]
printVerbose(">>> %s k-factor = %.2f" % (name0, kfactor),
verbosity,
level=2)
# check if sample list of contains to-be-stitched-sample
if len(stitch_list) < 2:
print warning("Could not stitch %s: less than two %s samples" %
(name0, name0))
for s in stitch_list:
print ">>> %s" % s.label
return sample_list
fill = max([len(s.label) for s in stitch_list])
name = kwargs.get('name', stitch_list[0].label)
# set renormalization scales with effective luminosity
for sample in stitch_list:
N_tot = sample.N
sigma = sample.sigma
N = N_tot
if name_incl in sample.filename:
N_incl = N_tot
elif not N_incl:
print warning("Could not stitch %s: N_incl == 0!" % name0)
return sample_list
else:
N = N_tot + N_incl * sigma / sigmaLO # effective luminosity
scale = lumi * kfactor * sigma * 1000 / N
weights.append("(NUP==%i ? %s : 1)" % (len(weights), scale))
printVerbose(
">>> stitching %s with scale %5.2f and cross section %7.2f pb" %
(sample.label.ljust(fill), scale, sigma),
verbosity,
level=2)
#print ">>> weight.append(%s)" % weights[-1]
if name_incl in sample.filename:
sample.scale = 1.0 # apply renormalization scale with weights
else:
sample.scale = scale # apply renormalization scale
# set weight of inclusive sample
for sample in stitch_list:
if sample.isPartOf(name_incl, *labels):
sample.scale = 1.0
sample.addWeight("*".join(weights))
# merge
mergeSamples(sample_list,
name0,
labels=labels,
name=name,
verbosity=verbosity)
