def morph_hist(self, file, directory, name, lower, upper, value) :
"""
Load histograms with name NAME that correspond to upper and lower bound,
determine morphed histogram corresponding to VALUE, and write the morphed
histogram to file.
"""
#print "loading fingerprint: ", file, directory, name
hist_lower = self.zero_safe(self.load_hist(file, directory, name.format(MASS=lower)))
hist_upper = self.zero_safe(self.load_hist(file, directory, name.format(MASS=upper)))
norm = self.norm_hist(hist_lower, hist_upper, float(lower), float(upper), float(value))
if self.trivial :
if abs(float(value)-float(lower)) < abs(float(upper)-float(value)) :
hist_morph = hist_lower.Clone(name.format(MASS=value)); hist_morph.SetTitle(name.format(MASS=value)); hist_morph.Scale(norm/hist_morph.Integral())
else :
hist_morph = hist_upper.Clone(name.format(MASS=value)); hist_morph.SetTitle(name.format(MASS=value)); hist_morph.Scale(norm/hist_morph.Integral())
else :
hist_morph = th1fmorph(name.format(MASS=value),name.format(MASS=value),hist_lower, hist_upper, float(lower), float(upper), float(value), norm, 0)
# th1fmorph() will set a value null if you are right on top of it
if not hist_lower and lower == value:
hist_lower = hist_morph
if not hist_upper and upper == value:
hist_upper = hist_morph
if self.verbose :
print "writing morphed histogram to file: name =", hist_morph.GetName(), "integral =[ %.5f | %.5f | %.5f ]" % (hist_lower.Integral(), hist_morph.Integral(), hist_upper.Integral())
if directory == "" :
file.cd()
else :
file.cd(directory)
hist_morph.Write(hist_morph.GetName())
def morph_hist(self, file, directory, name, lower, upper, value):
"""
Load histograms with name NAME that correspond to upper and lower bound,
determine morphed histogram corresponding to VALUE, and write the morphed
histogram to file.
"""
#print "loading fingerprint: ", file, directory, name
hist_lower = self.zero_safe(
self.load_hist(file, directory, name.format(MASS=lower)))
hist_upper = self.zero_safe(
self.load_hist(file, directory, name.format(MASS=upper)))
norm = self.norm_hist(hist_lower, hist_upper, float(lower),
float(upper), float(value))
hist_morph = th1fmorph(name.format(MASS=value),
name.format(MASS=value), hist_lower, hist_upper,
float(lower), float(upper), float(value), norm,
0)
# th1fmorph() will set a value null if you are right on top of it
if not hist_lower and lower == value:
hist_lower = hist_morph
if not hist_upper and upper == value:
hist_upper = hist_morph
if self.verbose:
print "writing morphed histogram to file: name =", hist_morph.GetName(
), "integral =[ %.5f | %.5f | %.5f ]" % (hist_lower.Integral(),
hist_morph.Integral(),
hist_upper.Integral())
if directory == "":
file.cd()
else:
file.cd(directory)
hist_morph.Write(hist_morph.GetName())
def morph(name, title, x, hist1, x1, hist2, x2):
'''
Given to histograms, <hist1> and <hist2>, which correspond to the parameters
<x1> and <x2>, return a TH1 with the given name and title, morphed to
correspond to parameter <x>.
The total yield for the target histogram is interpolated between the two
base histograms.
'''
yield1 = hist1.Integral()
yield2 = hist2.Integral()
interp_yield = interpolate(x1, yield1, x2, yield2, x)
return th1fmorph(name, title, hist1, hist2, x1, x2, x, interp_yield, 0)
def morph(name, title, x, hist1, x1, hist2, x2):
'''
Given to histograms, <hist1> and <hist2>, which correspond to the parameters
<x1> and <x2>, return a TH1 with the given name and title, morphed to
correspond to parameter <x>.
The total yield for the target histogram is interpolated between the two
base histograms.
'''
yield1 = hist1.Integral()
yield2 = hist2.Integral()
interp_yield = interpolate(x1, yield1, x2, yield2, x)
return th1fmorph(name, title, hist1, hist2, x1, x2, x, interp_yield, 0)
def single_template(self, dir, proc, mass, label, scale, MODE='MORPHED', debug=False) :
"""
Return a single template histogram for a given dir, proc, mass and label. The histogram will be scaled by scale
(corresponding to the cross section times BR of the corresponding Higgs boson). Scale will be modified by a linear
interpolation scale taking into account differences in acceptance and reconstruction efficiency as a function of
mass. Two modes exist to determine the template: MORPHED - will use horizontal template morphing (the morphing will
always be applied from the pivotal, which is closest to mass to minimize uncertainties from the interplation);
NEAREST_NEIGHBOUR - will use the closest mass point in the list of pivotals w/o any additional horizontal inter-
polation.
"""
## window of closest pivotal masses below/above mass. Window can be None, if no pivotals exist for a given dir. In
## this case return None
if self.ana_type=="Htaunu" and mass=="" :
single_template = self.load_hist(dir+'/'+proc+label).Clone(proc+label+'_template'); single_template.Scale(scale)
return single_template
window = self.pivotal_mass_window(float(mass), self.pivotals[dir])
if not window :
return None
if float(window[0]) == float(mass) and float(mass) == float(window[1]) :
## exact match with pivotal: clone exact pivotal
single_template = self.load_hist(dir+'/'+proc+mass+label).Clone(proc+mass+label+'_template'); single_template.Scale(scale)
elif float(window[0]) > float(mass) :
## mass out of bounds of pivotals (too small)
single_template = self.load_hist(dir+'/'+proc+window[0]+label).Clone(proc+window[0]+label+'_template'); single_template.Scale(scale)
elif float(window[1]) < float(mass) :
## mass out of bounds of pivotals (too large)
single_template = self.load_hist(dir+'/'+proc+window[1]+label).Clone(proc+window[1]+label+'_template'); single_template.Scale(scale)
else :
## mass somewhere between pivotals: masses is the tuple of the embracing pivotals, histos is the tuple of
## corresponding template histograms. The closest pivotal to mass is the first element in each of the tuples,
## the further away pivotal is second.
if (float(mass) - float(window[0])) < (float(window[1]) - float(mass)) :
## lower bound pivotal closer to mass
masses = (float(window[0]),float(window[1]))
histos = (self.load_hist(dir+'/'+proc+window[0]+label),self.load_hist(dir+'/'+proc+window[1]+label))
else :
## upper bound pivotal closer to mass
masses = (float(window[1]),float(window[0]))
histos = (self.load_hist(dir+'/'+proc+window[1]+label),self.load_hist(dir+'/'+proc+window[0]+label))
scale*= self.interpolation_scale((float(masses[0]),histos[0].Integral()), (float(masses[1]),histos[1].Integral()), float(mass), debug)
if MODE == 'MORPHED' :
single_template = th1fmorph(proc+str(mass)+label+'_template', proc+mass+label, histos[0], histos[1], masses[0], masses[1], float(mass), scale*histos[0].Integral(), 0)
if MODE == 'NEAREST_NEIGHBOUR' :
single_template = histos[0].Clone(proc+str(mass)+label+'_template'); single_template.Scale(scale)
return single_template
def morph_hist(self, file, directory, name, lower, upper, value) :
"""
Load histograms with name NAME that correspond to upper and lower bound,
determine morphed histogram corresponding to VALUE, and write the morphed
histogram to file.
"""
#print "loading fingerprint: ", file, directory, name
hist_lower = self.zero_safe(self.load_hist(file, directory, name.format(MASS=lower)))
hist_upper = self.zero_safe(self.load_hist(file, directory, name.format(MASS=upper)))
norm = self.norm_hist(hist_lower, hist_upper, float(lower), float(upper), float(value))
if self.trivial :
if abs(float(value)-float(lower)) < abs(float(upper)-float(value)) :
hist_morph = hist_lower.Clone(name.format(MASS=value)); hist_morph.SetTitle(name.format(MASS=value)); hist_morph.Scale(norm/hist_morph.Integral())
else :
hist_morph = hist_upper.Clone(name.format(MASS=value)); hist_morph.SetTitle(name.format(MASS=value)); hist_morph.Scale(norm/hist_morph.Integral())
else :
hist_morph = th1fmorph(name.format(MASS=value),name.format(MASS=value),hist_lower, hist_upper, float(lower), float(upper), float(value), norm, 0)
# th1fmorph() will set a value null if you are right on top of it
if not hist_lower and lower == value:
hist_lower = hist_morph
if not hist_upper and upper == value:
hist_upper = hist_morph
if self.verbose :
if not "_fine_binning" in name:
print "writing morphed histogram to file: name =", hist_morph.GetName(), "integral =[ %.5f | %.5f | %.5f ]" % (hist_lower.Integral(), hist_morph.Integral(), hist_upper.Integral())
else:
print "writing morphed '_fine_binning' histogram to file: name =", hist_morph.GetName().replace("_fine_binning",""), "integral =[ %.5f | %.5f | %.5f ]" % (hist_lower.Integral(), hist_morph.Integral(), hist_upper.Integral())
if directory == "" :
file.cd()
else :
file.cd(directory)
if "_fine_binning" in name:
#rebin to the usual binning using the histogram without the '_fine_binning' label
self.rebin_hist(hist_morph,self.load_hist(file, directory, hist_lower.GetName().replace("_fine_binning","").format(MASS=lower)))
#save removing 'fine_binning' from the name
hist_morph.Write(hist_morph.GetName().replace("_fine_binning",""))
else :
hist_morph.Write(hist_morph.GetName())
def main(args, opt):
os.system('mkdir -p %s'%opt.outDir)
mcfiles = {} # procname -> filename
datafiles = {} # procname -> filename
try:
for fname in os.listdir(args[0]):
if not osp.splitext(fname)[1] == '.root': continue
isdata,procname,splitno = resolveFilename(fname)
if isdata:
if not procname in datafiles:
datafiles[procname] = []
datafiles[procname].append(osp.join(args[0],fname))
else:
if 'QCD' in procname: continue ## exclude QCD
if procname == 'TTJets_MSDecays_172v5': continue ## have those already
if 'SingleT' in procname: continue ## have those already
if not procname in mcfiles:
mcfiles[procname] = []
mcfiles[procname].append(osp.join(args[0],fname))
except IndexError:
print "Please provide a valid input directory"
exit(-1)
## Produce (or read) the histogram data
bghistos = makeBackgroundHistos(mcfiles, opt)
cachefile = open(".xsecweights.pck", 'r')
xsecweights = pickle.load(cachefile)
cachefile.close()
print '>>> Read xsec weights from cache (.xsecweights.pck)'
cachefile = open(".svldyscalefactors.pck", 'r')
dySFs = pickle.load(cachefile)
cachefile.close()
print '>>> Read DY scale factors from cache (.svldyscalefactors.pck)'
cachefile = open(".svlqcdtemplates.pck", 'r')
qcdTemplates = pickle.load(cachefile)
cachefile.close()
print '>>> Read QCD templates from cache (.svlqcdtemplates.pck)'
## Read SV Track multiplicity weights:
from extractNtrkWeights import extractNTrkWeights
ntkWeights = extractNTrkWeights()
## Now add them up with proper scales
mcprocesses = [k for k in mcfiles.keys() if not 'Data8TeV' in k]
bghistos_added = sumBGHistos(processes=mcprocesses,
bghistos=bghistos,
xsecweights=xsecweights,
ntkWeights=ntkWeights,
dySFs=dySFs,
qcdTemplates=qcdTemplates,
opt=opt)
bghistos_added_dyup = sumBGHistos(processes=mcprocesses,
bghistos=bghistos,
xsecweights=xsecweights,
ntkWeights=ntkWeights,
dySFs=dySFs,
qcdTemplates=qcdTemplates,
opt=opt,
dyScale=1.3)
bghistos_added_dydn = sumBGHistos(processes=mcprocesses,
bghistos=bghistos,
xsecweights=xsecweights,
ntkWeights=ntkWeights,
dySFs=dySFs,
qcdTemplates=qcdTemplates,
opt=opt,
dyScale=0.7)
bghistos_added_qcdup = sumBGHistos(processes=mcprocesses,
bghistos=bghistos,
xsecweights=xsecweights,
ntkWeights=ntkWeights,
dySFs=dySFs,
qcdTemplates=qcdTemplates,
opt=opt,
qcdScale=1.1)
bghistos_added_qcddn = sumBGHistos(processes=mcprocesses,
bghistos=bghistos,
xsecweights=xsecweights,
ntkWeights=ntkWeights,
dySFs=dySFs,
qcdTemplates=qcdTemplates,
opt=opt,
qcdScale=0.9)
## Produce data histograms
datahistos = makeDataHistos(datafiles, opt)
datahistos_added = sumDataHistos(datafiles.keys(), datahistos)
# Rebin also data, if required:
if opt.rebin>0:
for hist in datahistos_added.values():
hist.Rebin(opt.rebin)
## Save the background only shapes separately as templates for the fit
cachefile = open(".svlbgtemplates.pck", 'w')
pickle.dump(bghistos_added, cachefile, pickle.HIGHEST_PROTOCOL)
print '>>> Dumped bg templates to cache (.svlbgtemplates.pck)'
cachefile.close()
## Read syst histos:
cachefile = open(".svlsysthistos.pck", 'r')
systhistos = pickle.load(cachefile)
print '>>> Read systematics histograms from cache (.svlsysthistos.pck)'
cachefile.close()
## Read mass scan histos:
cachefile = open(".svlmasshistos.pck", 'r')
masshistos = pickle.load(cachefile)
print '>>> Read mass scan histograms from cache (.svlmasshistos.pck)'
# (tag, chan, mass, comb) -> histo
# (tag, chan, mass, comb, ntk) -> histo
cachefile.close()
ofi = ROOT.TFile.Open(osp.join(opt.outDir,'pe_inputs.root'),'RECREATE')
ofi.cd()
#####################################################
## Central mass point and syst samples
for syst in ([s for s,_,_,_ in ALLSYSTS] +
['dyup','dydown','qcdup','qcddown','ntkmult']):
odir = ofi.mkdir(syst + '_172v5')
odir.cd()
for tag,_,_ in SELECTIONS:
for ntk,_ in NTRKBINS:
hname = "SVLMass_%s_%s_%s" % (tag,syst+'_172v5',ntk)
if not syst in ['dyup','dydown','qcdup','qcddown','ntkmult',
'tchscaleup','tchscaledown',
'twchscaleup','twchscaledown']:
hfinal = systhistos[(tag,syst,'tot',ntk)].Clone(hname)
else:
hfinal = systhistos[(tag,'nominal','tot',ntk)].Clone(hname)
try:
## Systs from separate samples
if syst in ['tchscaleup','tchscaledown',
'twchscaleup','twchscaledown']:
scale = LUMI*xsecweights[CHANMASSTOPROCNAME[('tt', 172.5)]]
else:
scale = LUMI*xsecweights[SYSTTOPROCNAME[syst][0]]
except KeyError:
## Systs from event weights
scale = LUMI*xsecweights[CHANMASSTOPROCNAME[('tt', 172.5)]]
hfinal.Scale(scale)
## Renormalize some variations with event weights
if syst in SYSTSTOBERENORMALIZED:
normintegral = systhistos[(tag,'nominal','tot',ntk)].Integral()
normintegral *= LUMI*xsecweights[CHANMASSTOPROCNAME[('tt', 172.5)]]
normintegral /= hfinal.Integral()
hfinal.Scale(normintegral)
## Add single top
stProcs=['t', 'tbar', 'tW', 'tbarW']
stSystProcs=[]
if 'tchscale' in syst:
stProcs=['tW', 'tbarW']
stSystProcs=['t', 'tbar']
if 'twchscale' in syst:
stProcs=['t', 'tbar']
stSystProcs=['tW', 'tbarW']
for st in stProcs:
hsinglet = masshistos[(tag, st, 172.5,'tot',ntk)].Clone('%s_%s'%(hname,st))
hsinglet.Scale(LUMI*xsecweights[CHANMASSTOPROCNAME[(st, 172.5)]])
hfinal.Add(hsinglet)
for st in stSystProcs:
hsinglet = systhistos[(tag, syst, 'tot', ntk)].Clone('%s_%s'%(hname,st))
hsinglet.Scale(LUMI*xsecweights[CHANMASSTOPROCNAME[(st, 172.5)]])
hfinal.Add(hsinglet)
## Add the backgrounds
if not syst in ['dyup','dydown','qcdup','qcddown']:
hfinal.Add(bghistos_added[(tag,ntk)])
else: ## From the scaled bghistos if necessary
bghistos_added_scaled = {
'dyup' : bghistos_added_dyup,
'dydown' : bghistos_added_dydn,
'qcdup' : bghistos_added_qcdup,
'qcddown' : bghistos_added_qcddn,
}[syst]
hfinal.Add(bghistos_added_scaled[(tag,ntk)])
## Rebin if requested
if opt.rebin>0:
hfinal.Rebin(opt.rebin)
## Scale by SV track multiplicity weights:
if not syst == 'ntkmult':
hfinal.Scale(ntkWeights['inclusive'][ntk])
## Write out to file
hfinal.Write(hname, ROOT.TObject.kOverwrite)
#####################################################
## Non-central mass points
ROOT.gSystem.Load('libUserCodeTopMassSecVtx.so')
from ROOT import th1fmorph
# extract mass points from dictionary
mass_points = sorted(list(set([key[2] for key in masshistos.keys()])))
mass_points = mass_points[1:-1] # remove outermost points
debughistos = []
for mass in mass_points:
if mass == 172.5: continue
mname = 'nominal_%s' % str(mass).replace('.','v')
odir = ofi.mkdir(mname)
odir.cd()
for tag,_,_ in SELECTIONS:
for ntk,_ in NTRKBINS:
hname = "SVLMass_%s_%s_%s" % (tag,mname,ntk)
hfinal = masshistos[(tag,'tt',mass,'tot',ntk)].Clone(hname)
hfinal.Scale(LUMI*xsecweights[CHANMASSTOPROCNAME[('tt', mass)]])
## Add single top (t-channel, for which we have the samples)
for st in ['t', 'tbar']:
hsinglet = masshistos[(tag, st, mass,'tot',ntk)].Clone('%s_%s'%(hname,st))
hsinglet.Scale(LUMI*xsecweights[CHANMASSTOPROCNAME[(st, mass)]])
hfinal.Add(hsinglet)
## Add single top (tW-channel, for which we don't have samples)
## Morph between the two extreme mass points to get
## the non existing ones
for st in ['tW', 'tbarW']:
if mass not in [166.5, 178.5]:
hsingletW = th1fmorph('%s_%s_morph'%(hname,st),
'%s_%s_morphed'%(hname,st),
masshistos[(tag, 'tW', 166.5,'tot',ntk)],
masshistos[(tag, 'tW', 178.5,'tot',ntk)],
166.5, 178.5, mass,
masshistos[(tag, 'tW', 166.5,'tot',ntk)].Integral())
hsingletW.Scale(LUMI*xsecweights[CHANMASSTOPROCNAME[(st, 166.5)]]
* TWXSECS[mass]/TWXSECS[166.5])
hsingletW.SetDirectory(0)
else:
hsingletW = masshistos[(tag, st, mass,'tot',ntk)].Clone('%s_%s'%(hname,st))
hsingletW.Scale(LUMI*xsecweights[CHANMASSTOPROCNAME[(st, mass)]])
hfinal.Add(hsingletW)
## Add the combined backgrounds
hfinal.Add(bghistos_added[(tag,ntk)])
## Rebin if requested
if opt.rebin>0:
hfinal.Rebin(opt.rebin)
## Scale by SV track multiplicity weights:
hfinal.Scale(ntkWeights['inclusive'][ntk])
## Write out to file
hfinal.Write(hname, ROOT.TObject.kOverwrite)
## Write also data histos
ofi.cd()
odir = ofi.mkdir('data')
odir.cd()
for tag,_,_ in SELECTIONS:
for ntk,_ in NTRKBINS:
hname = "SVLMass_%s_data_%s" % (tag,ntk)
datahistos_added[(tag,ntk)].Write(hname, ROOT.TObject.kOverwrite)
print ('>>> Wrote pseudo experiment inputs to file (%s)' %
osp.join(opt.outDir,'pe_inputs.root'))
ofi.Write()
ofi.Close()
return 0
def main(args, opt):
os.system('mkdir -p %s' % opt.outDir)
mcfiles = {} # procname -> filename
datafiles = {} # procname -> filename
try:
for fname in os.listdir(args[0]):
if not osp.splitext(fname)[1] == '.root': continue
isdata, procname, splitno = resolveFilename(fname)
if isdata:
if not procname in datafiles:
datafiles[procname] = []
datafiles[procname].append(osp.join(args[0], fname))
else:
if 'QCD' in procname: continue ## exclude QCD
if procname == 'TTJets_MSDecays_172v5':
continue ## have those already
if 'SingleT' in procname: continue ## have those already
if not procname in mcfiles:
mcfiles[procname] = []
mcfiles[procname].append(osp.join(args[0], fname))
except IndexError:
print "Please provide a valid input directory"
exit(-1)
## Produce (or read) the histogram data
bghistos = makeBackgroundHistos(mcfiles, opt)
cachefile = open(".xsecweights.pck", 'r')
xsecweights = pickle.load(cachefile)
cachefile.close()
print '>>> Read xsec weights from cache (.xsecweights.pck)'
cachefile = open(".svldyscalefactors.pck", 'r')
dySFs = pickle.load(cachefile)
cachefile.close()
print '>>> Read DY scale factors from cache (.svldyscalefactors.pck)'
cachefile = open(".svlqcdtemplates.pck", 'r')
qcdTemplates = pickle.load(cachefile)
cachefile.close()
print '>>> Read QCD templates from cache (.svlqcdtemplates.pck)'
## Read SV Track multiplicity weights:
from extractNtrkWeights import extractNTrkWeights
ntkWeights = extractNTrkWeights()
## Now add them up with proper scales
mcprocesses = [k for k in mcfiles.keys() if not 'Data8TeV' in k]
bghistos_added = sumBGHistos(processes=mcprocesses,
bghistos=bghistos,
xsecweights=xsecweights,
ntkWeights=ntkWeights,
dySFs=dySFs,
qcdTemplates=qcdTemplates,
opt=opt)
bghistos_added_dyup = sumBGHistos(processes=mcprocesses,
bghistos=bghistos,
xsecweights=xsecweights,
ntkWeights=ntkWeights,
dySFs=dySFs,
qcdTemplates=qcdTemplates,
opt=opt,
dyScale=1.3)
bghistos_added_dydn = sumBGHistos(processes=mcprocesses,
bghistos=bghistos,
xsecweights=xsecweights,
ntkWeights=ntkWeights,
dySFs=dySFs,
qcdTemplates=qcdTemplates,
opt=opt,
dyScale=0.7)
bghistos_added_qcdup = sumBGHistos(processes=mcprocesses,
bghistos=bghistos,
xsecweights=xsecweights,
ntkWeights=ntkWeights,
dySFs=dySFs,
qcdTemplates=qcdTemplates,
opt=opt,
qcdScale=1.1)
bghistos_added_qcddn = sumBGHistos(processes=mcprocesses,
bghistos=bghistos,
xsecweights=xsecweights,
ntkWeights=ntkWeights,
dySFs=dySFs,
qcdTemplates=qcdTemplates,
opt=opt,
qcdScale=0.9)
## Produce data histograms
datahistos = makeDataHistos(datafiles, opt)
datahistos_added = sumDataHistos(datafiles.keys(), datahistos)
# Rebin also data, if required:
if opt.rebin > 0:
for hist in datahistos_added.values():
hist.Rebin(opt.rebin)
## Save the background only shapes separately as templates for the fit
cachefile = open(".svlbgtemplates.pck", 'w')
pickle.dump(bghistos_added, cachefile, pickle.HIGHEST_PROTOCOL)
print '>>> Dumped bg templates to cache (.svlbgtemplates.pck)'
cachefile.close()
## Read syst histos:
cachefile = open(".svlsysthistos.pck", 'r')
systhistos = pickle.load(cachefile)
print '>>> Read systematics histograms from cache (.svlsysthistos.pck)'
cachefile.close()
## Read mass scan histos:
cachefile = open(".svlmasshistos.pck", 'r')
masshistos = pickle.load(cachefile)
print '>>> Read mass scan histograms from cache (.svlmasshistos.pck)'
# (tag, chan, mass, comb) -> histo
# (tag, chan, mass, comb, ntk) -> histo
cachefile.close()
## Signal only (tt+t+tW) shapes
signalonly = {}
ofi = ROOT.TFile.Open(osp.join(opt.outDir, 'pe_inputs.root'), 'RECREATE')
ofi.cd()
#####################################################
## Central mass point and syst samples
to_be_processed = ([s for s, _, _, _ in ALLSYSTS] +
['dyup', 'dydown', 'qcdup', 'qcddown', 'ntkmult'])
if opt.skip_systs: to_be_processed = ['nominal']
for syst in to_be_processed:
odir = ofi.mkdir(syst + '_172v5')
odir.cd()
for tag, _, _ in SELECTIONS:
for ntk, _ in NTRKBINS:
hname = "SVLMass_%s_%s_%s" % (tag, syst + '_172v5', ntk)
if not syst in [
'dyup', 'dydown', 'qcdup', 'qcddown', 'ntkmult',
'tchscaleup', 'tchscaledown', 'twchscaleup',
'twchscaledown'
]:
hfinal = systhistos[(tag, syst, 'tot', ntk)].Clone(hname)
else:
hfinal = systhistos[(tag, 'nominal', 'tot',
ntk)].Clone(hname)
try:
## Systs from separate samples
if syst in [
'tchscaleup', 'tchscaledown', 'twchscaleup',
'twchscaledown'
]:
scale = LUMI * xsecweights[CHANMASSTOPROCNAME[('tt',
172.5)]]
else:
scale = LUMI * xsecweights[SYSTTOPROCNAME[syst][0]]
except KeyError:
## Systs from event weights
scale = LUMI * xsecweights[CHANMASSTOPROCNAME[('tt',
172.5)]]
hfinal.Scale(scale)
## Renormalize some variations with event weights
if syst in SYSTSTOBERENORMALIZED:
normintegral = systhistos[(tag, 'nominal', 'tot',
ntk)].Integral()
normintegral *= LUMI * xsecweights[CHANMASSTOPROCNAME[
('tt', 172.5)]]
normintegral /= hfinal.Integral()
hfinal.Scale(normintegral)
## Add single top
stProcs = ['t', 'tbar', 'tW', 'tbarW']
stSystProcs = []
if 'tchscale' in syst:
stProcs = ['tW', 'tbarW']
stSystProcs = ['t', 'tbar']
if 'twchscale' in syst:
stProcs = ['t', 'tbar']
stSystProcs = ['tW', 'tbarW']
for st in stProcs:
hsinglet = masshistos[(tag, st, 172.5, 'tot',
ntk)].Clone('%s_%s' % (hname, st))
hsinglet.Scale(
LUMI * xsecweights[CHANMASSTOPROCNAME[(st, 172.5)]])
hfinal.Add(hsinglet)
for st in stSystProcs:
hsinglet = systhistos[(tag, syst, 'tot',
ntk)].Clone('%s_%s' % (hname, st))
hsinglet.Scale(
LUMI * xsecweights[CHANMASSTOPROCNAME[(st, 172.5)]])
hfinal.Add(hsinglet)
## Save signal only shapes
if syst == 'nominal':
signalonly[(tag, 172.5,
ntk)] = hfinal.Clone('%s_sigonly' % hname)
signalonly[(tag, 172.5,
ntk)].Scale(ntkWeights['inclusive'][ntk])
if opt.rebin > 0:
signalonly[(tag, 172.5, ntk)].Rebin(opt.rebin)
## Add the backgrounds
if not syst in ['dyup', 'dydown', 'qcdup', 'qcddown']:
hfinal.Add(bghistos_added[(tag, ntk)])
else: ## From the scaled bghistos if necessary
bghistos_added_scaled = {
'dyup': bghistos_added_dyup,
'dydown': bghistos_added_dydn,
'qcdup': bghistos_added_qcdup,
'qcddown': bghistos_added_qcddn,
}[syst]
hfinal.Add(bghistos_added_scaled[(tag, ntk)])
## Rebin if requested
if opt.rebin > 0:
hfinal.Rebin(opt.rebin)
## Scale by SV track multiplicity weights:
if not syst == 'ntkmult':
hfinal.Scale(ntkWeights['inclusive'][ntk])
## Write out to file
hfinal.Write(hname, ROOT.TObject.kOverwrite)
#####################################################
## Non-central mass points
ROOT.gSystem.Load('libUserCodeTopMassSecVtx.so')
from ROOT import th1fmorph
# extract mass points from dictionary
mass_points = sorted(list(set([key[2] for key in masshistos.keys()])))
mass_points = mass_points[1:-1] # remove outermost points
debughistos = []
for mass in mass_points:
if mass == 172.5: continue
mname = 'nominal_%s' % str(mass).replace('.', 'v')
odir = ofi.mkdir(mname)
odir.cd()
for tag, _, _ in SELECTIONS:
for ntk, _ in NTRKBINS:
hname = "SVLMass_%s_%s_%s" % (tag, mname, ntk)
hfinal = masshistos[(tag, 'tt', mass, 'tot', ntk)].Clone(hname)
hfinal.Scale(LUMI *
xsecweights[CHANMASSTOPROCNAME[('tt', mass)]])
## Add single top (t-channel, for which we have the samples)
for st in ['t', 'tbar']:
hsinglet = masshistos[(tag, st, mass, 'tot',
ntk)].Clone('%s_%s' % (hname, st))
hsinglet.Scale(LUMI *
xsecweights[CHANMASSTOPROCNAME[(st, mass)]])
hfinal.Add(hsinglet)
## Add single top (tW-channel, for which we don't have samples)
## Morph between the two extreme mass points to get
## the non existing ones
for st in ['tW', 'tbarW']:
if mass not in [166.5, 178.5]:
hsingletW = th1fmorph(
'%s_%s_morph' % (hname, st),
'%s_%s_morphed' % (hname, st),
masshistos[(tag, 'tW', 166.5, 'tot', ntk)],
masshistos[(tag, 'tW', 178.5, 'tot', ntk)], 166.5,
178.5, mass, masshistos[(tag, 'tW', 166.5, 'tot',
ntk)].Integral())
hsingletW.Scale(
LUMI *
xsecweights[CHANMASSTOPROCNAME[(st, 166.5)]] *
TWXSECS[mass] / TWXSECS[166.5])
hsingletW.SetDirectory(0)
else:
hsingletW = masshistos[(tag, st, mass, 'tot',
ntk)].Clone('%s_%s' %
(hname, st))
hsingletW.Scale(
LUMI * xsecweights[CHANMASSTOPROCNAME[(st, mass)]])
hfinal.Add(hsingletW)
## Save signal only shapes
signalonly[(tag, mass,
ntk)] = hfinal.Clone('%s_sigonly' % hname)
signalonly[(tag, mass,
ntk)].Scale(ntkWeights['inclusive'][ntk])
if opt.rebin > 0:
signalonly[(tag, mass, ntk)].Rebin(opt.rebin)
## Add the combined backgrounds
hfinal.Add(bghistos_added[(tag, ntk)])
## Rebin if requested
if opt.rebin > 0:
hfinal.Rebin(opt.rebin)
## Scale by SV track multiplicity weights:
hfinal.Scale(ntkWeights['inclusive'][ntk])
## Write out to file
hfinal.Write(hname, ROOT.TObject.kOverwrite)
## Save the signal only shapes (tt+t+tW) as input for the combined plot
cachefile = open(".svlsignalshapes.pck", 'w')
pickle.dump(signalonly, cachefile, pickle.HIGHEST_PROTOCOL)
print '>>> Dumped signal only shapes to cache (.svlsignalshapes.pck)'
cachefile.close()
## Write also data histos
ofi.cd()
odir = ofi.mkdir('data')
odir.cd()
for tag, _, _ in SELECTIONS:
for ntk, _ in NTRKBINS:
hname = "SVLMass_%s_data_%s" % (tag, ntk)
datahistos_added[(tag, ntk)].Write(hname, ROOT.TObject.kOverwrite)
print('>>> Wrote pseudo experiment inputs to file (%s)' %
osp.join(opt.outDir, 'pe_inputs.root'))
ofi.Write()
ofi.Close()
return 0
def single_template(self,
dir,
proc,
mass,
label,
scale,
MODE='MORPHED',
debug=False):
"""
Return a single template histogram for a given dir, proc, mass and label. The histogram will be scaled by scale
(corresponding to the cross section times BR of the corresponding Higgs boson). Scale will be modified by a linear
interpolation scale taking into account differences in acceptance and reconstruction efficiency as a function of
mass. Two modes exist to determine the template: MORPHED - will use horizontal template morphing (the morphing will
always be applied from the pivotal, which is closest to mass to minimize uncertainties from the interplation);
NEAREST_NEIGHBOUR - will use the closest mass point in the list of pivotals w/o any additional horizontal inter-
polation.
"""
## window of closest pivotal masses below/above mass. Window can be None, if no pivotals exist for a given dir. In
## this case return None
if self.ana_type == "Htaunu" and mass == "":
single_template = self.load_hist(dir + '/' + proc +
label).Clone(proc + label +
'_template')
single_template.Scale(scale)
return single_template
window = self.pivotal_mass_window(float(mass), self.pivotals[dir])
if not window:
return None
if float(window[0]) == float(mass) and float(mass) == float(window[1]):
## exact match with pivotal: clone exact pivotal
single_template = self.load_hist(dir + '/' + proc + mass +
label).Clone(proc + mass + label +
'_template')
single_template.Scale(scale)
elif float(window[0]) > float(mass):
## mass out of bounds of pivotals (too small)
single_template = self.load_hist(dir + '/' + proc + window[0] +
label).Clone(proc + window[0] +
label + '_template')
single_template.Scale(scale)
elif float(window[1]) < float(mass):
## mass out of bounds of pivotals (too large)
single_template = self.load_hist(dir + '/' + proc + window[1] +
label).Clone(proc + window[1] +
label + '_template')
single_template.Scale(scale)
else:
## mass somewhere between pivotals: masses is the tuple of the embracing pivotals, histos is the tuple of
## corresponding template histograms. The closest pivotal to mass is the first element in each of the tuples,
## the further away pivotal is second.
if (float(mass) - float(window[0])) < (float(window[1]) -
float(mass)):
## lower bound pivotal closer to mass
masses = (float(window[0]), float(window[1]))
histos = (self.load_hist(dir + '/' + proc + window[0] + label),
self.load_hist(dir + '/' + proc + window[1] + label))
else:
## upper bound pivotal closer to mass
masses = (float(window[1]), float(window[0]))
histos = (self.load_hist(dir + '/' + proc + window[1] + label),
self.load_hist(dir + '/' + proc + window[0] + label))
scale *= self.interpolation_scale(
(float(masses[0]), histos[0].Integral()),
(float(masses[1]), histos[1].Integral()), float(mass), debug)
if MODE == 'MORPHED':
single_template = th1fmorph(
proc + str(mass) + label + '_template',
proc + mass + label, histos[0], histos[1], masses[0],
masses[1], float(mass), scale * histos[0].Integral(), 0)
if MODE == 'NEAREST_NEIGHBOUR':
single_template = histos[0].Clone(proc + str(mass) + label +
'_template')
single_template.Scale(scale)
return single_template