def create_samples(self):
print "total samples:",len(sample_names)
for index in range(len(sample_names)):
sample_name = sample_names[index]
sample = Sample(sample_name, sample_colors[index])
file_name = input_dir+sample_name+"_combined.root"
yields_dic = self.read_hist(file_name)
print "Sample:", sample_name+";",
if "data" in sample_name:
sample.setData()
self.data_sample = sample
for region in regions:
nevts, nerror = yields_dic[region]
sample.buildHisto([nevts], region, "cuts", 0.5)
print region,str(round(nevts,3))+";",
print
continue
sample.setNormByTheory()
for region in regions:
nevts, nerror = yields_dic[region]
nevts *= weight
nerror *= weight
if "Dijets" in sample_name and "SR" in region:
if self.cut == 10:
nevts = 4.07
nerror = math.sqrt(nevts)
if self.cut == 14:
nevts = 2.42
nerror = math.sqrt(nevts)
sample.buildHisto([nevts], region, "cuts", 0.5)
sample.buildStatErrors([nerror], region, "cuts")
print region,str(round(nevts,3))+";",
print
#sample.setStatConfig(True)
sample.setFileList([in_file_path])
## add systematic??
sample.addSystematic(Systematic(sample_name+"_stats",\
configMgr.weights, 1.2, 0.8, "user", "userOverallSys"))
#for systematic in self.sys_common:
#sample.addSystematic(systematic)
self.set_norm_factor(sample)
z_sample.setNormFactor("mu_z", 1, 0, 100)
z_sample.setStatConfig(use_stat)
sample_list_bkg.append(z_sample)
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# ttbar
ttbar_sample = Sample("ttbar", kGreen+2)
ttbar_sample.setNormFactor("mu_ttbar", 1, 0, 100)
ttbar_sample.setStatConfig(use_stat)
sample_list_bkg.append(ttbar_sample)
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# data
data_sample = Sample("data",kBlack)
data_sample.setData()
sample_list_data.append(data_sample)
# - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
# set the file from which the samples should be taken
# for sl in itertools.chain(sample_list_bkg, sample_list_data):
for sl in sample_list_bkg:
sl.setFileList(bkg_files)
for sl in sample_list_data:
sl.setFileList(data_files)
# ------------------------------------------------------------------------------
# add systematics to each of the samples
def addSystematic(sample_list, syst_list):
)
# topKtScale = Systematic("KtScaleTop",configMgr.weights,ktScaleTopHighWeights,ktScaleTopLowWeights,"weight","normHistoSys")
# JES (tree-based)
jes = Systematic("JES", "_NoSys", "_JESup", "_JESdown", "tree", "overallSys")
configMgr.nomName = "_NoSys"
# -------------------------------------------
# List of samples and their plotting colours
# -------------------------------------------
topSample = Sample("Top", kGreen - 9)
# topSample.setNormFactor("mu_Top",1.,0.,5.)
wzSample = Sample("WZ", kAzure + 1)
# wzSample.setNormFactor("mu_WZ",1.,0.,5.)
dataSample = Sample("Data", kBlack)
dataSample.setData()
# **************
# Exclusion fit
# **************
# Fit config instance
exclusionFitConfig = configMgr.addTopLevelXML("Exclusion")
meas = exclusionFitConfig.addMeasurement(name="NormalMeasurement", lumi=1.0, lumiErr=0.039)
meas.addPOI("mu_SIG")
# Samples
exclusionFitConfig.addSamples([topSample, wzSample, dataSample])
# Systematics
for region_name, region_infile in zip(region_names, region_infiles):
binvals, edges = HistogramImporter.import_histogram(
os.path.join(indir, region_infile), template_name)
bin_width = edges[1] - edges[0]
cur_sample.buildHisto(binvals,
region_name,
"mBB",
binLow=edges[0],
binWidth=bin_width)
samples.append(cur_sample)
# also make the (Asimov) data sample
data_sample = Sample("data", ROOT.kBlack)
data_sample.setData()
# in each region, it holds the total event content
for region_name, region_infile in zip(region_names, region_infiles):
binvals = None
for sample_name, template_name in zip(sample_names, template_names):
sample_binvals, edges = HistogramImporter.import_histogram(
os.path.join(indir, region_infile), template_name)
bin_width = edges[1] - edges[0]
if not binvals:
binvals = sample_binvals
else:
binvals = [
binval + sample_binval
configMgr.weights = "1."
# Define samples
bkgSample = Sample("Bkg", ROOT.kGreen - 9)
bkgSample.setStatConfig(True)
bkgSample.buildHisto([nbkg], "SR", "cuts", 0.5)
bkgSample.addSystematic(ucb)
sigSample = Sample("GGM_GG_bhmix_%d_%d" % (args.m3, args.mu), ROOT.kOrange + 3)
sigSample.setNormFactor("mu_SIG", 1., 0., 10.)
#sigSample.setStatConfig(True)
sigSample.setNormByTheory()
sigSample.buildHisto([nsig], "SR", "cuts", 0.5)
dataSample = Sample("Data", ROOT.kBlack)
dataSample.setData()
dataSample.buildHisto([ndata], "SR", "cuts", 0.5)
# Define top-level
ana = configMgr.addFitConfig("Disc")
ana.addSamples([bkgSample, sigSample, dataSample])
ana.setSignalSample(sigSample)
# Define measurement
meas = ana.addMeasurement(name="NormalMeasurement",
lumi=1.0,
lumiErr=lumiError)
meas.addPOI("mu_SIG")
meas.addParamSetting("Lumi", True)
# Add the channel
diphotons.setNormByTheory()
diphotons.setStatConfig(True)
diphotons.addSystematic(diphotonsNorm)
st = Sample("st",8) # dark green
st.setNormByTheory()
st.setStatConfig(True)
st.addSystematic(stNorm)
gj = Sample("gj",28) # brown
gj.setStatConfig(True)
gj.addSystematic(qcdElNorm)
#gj.setQCD()
data = Sample("data",kBlack)
data.setData()
commonSamples = [ttbargamma, Wgamma, Wjets, ttbarDilep, ttbarLepjets, st, Zgamma, Zleplep, diboson, data]
## Parameters of the Measurement
measName = "BasicMeasurement"
measLumi = 1.
measLumiError = 0.039
## Parameters of Channels
cutsNBins = 1
cutsBinLow = 0.0
cutsBinHigh = 1.0
## Bkg-only fit (add the common systematics, which also affect the signal when cloning)
bkgOnly = configMgr.addTopLevelXML("LepPhoton_BkgOnly")
configMgr.histCacheFile = "data/" + configMgr.analysisName + ".root"
configMgr.outputFileName = "results/" + configMgr.analysisName + "_Output.root"
print "is discovery ? %s" % (myFitType == FitType.Discovery)
sample_bkg0 = Sample("bkg0", ROOT.kBlue)
sample_bkg0.setStatConfig(True)
#sample_bkg1 = Sample("bkg1", ROOT.kGreen)
#sample_bkg1.setStatConfig(True)
#
#sample_bkg2 = Sample("bkg2", ROOT.kMagenta)
#sample_bkg2.setStatConfig(True)
sample_data = Sample("data", ROOT.kBlack)
sample_data.setData()
#sample_sig = Sample("sig", ROOT.kRed)
#sample_sig.setStatConfig(True)
all_samples = [sample_bkg0,
sample_data] #, sample_bkg1, sample_bkg2, sample_data]
# systematics
syst_bkg0_cr = Systematic("SYST_BKG_CR", configMgr.weights, 1.0 + 0.10,
1.0 - 0.10, "user", "userHistoSys")
syst_bkg0_sr = Systematic("SYST_BKG_SR", configMgr.weights, 1.0 + 0.05,
1.0 - 0.05, "user", "userHistoSys")
sample_bkg0.addSystematic(syst_bkg0_sr)
#norm_syst_bkg1 = Systematic("Norm_Bkg1_0", configMgr.weights, 1.0 + 0.1, 1.0 - 0.1, "user", "userHistoSys")
def common_setting(mass):
from configManager import configMgr
from ROOT import kBlack, kGray, kRed, kPink, kViolet, kBlue, kAzure, kGreen, \
kOrange
from configWriter import Sample
from systematic import Systematic
import os
color_dict = {
"Zbb": kAzure,
"Zbc": kAzure,
"Zbl": kAzure,
"Zcc": kAzure,
"Zcl": kBlue,
"Zl": kBlue,
"Wbb": kGreen,
"Wbc": kGreen,
"Wbl": kGreen,
"Wcc": kGreen,
"Wcl": kGreen,
"Wl": kGreen,
"ttbar": kOrange,
"stop": kOrange,
"stopWt": kOrange,
"ZZPw": kGray,
"WZPw": kGray,
"WWPw": kGray,
"fakes": kPink,
"Zjets": kAzure,
"Wjets": kGreen,
"top": kOrange,
"diboson": kGray,
"$Z\\tau\\tau$+HF": kAzure,
"$Z\\tau\\tau$+LF": kBlue,
"$W$+jets": kGreen,
"$Zee$": kViolet,
"Zhf": kAzure,
"Zlf": kBlue,
"Zee": kViolet,
"others": kViolet,
signal_prefix + "1000": kRed,
signal_prefix + "1100": kRed,
signal_prefix + "1200": kRed,
signal_prefix + "1400": kRed,
signal_prefix + "1600": kRed,
signal_prefix + "1800": kRed,
signal_prefix + "2000": kRed,
signal_prefix + "2500": kRed,
signal_prefix + "3000": kRed,
# Add your new processes here
"VH": kGray + 2,
"VHtautau": kGray + 2,
"ttH": kGray + 2,
}
##########################
# Setting the parameters of the hypothesis test
configMgr.doExclusion = True # True=exclusion, False=discovery
configMgr.nTOYs = 10000 # default=5000
configMgr.calculatorType = 0 # 2=asymptotic calculator, 0=frequentist calculator
configMgr.testStatType = 3 # 3=one-sided profile likelihood test statistic (LHC default)
configMgr.nPoints = 30 # number of values scanned of signal-strength for upper-limit determination of signal strength.
configMgr.writeXML = False
configMgr.seed = 40
configMgr.toySeedSet = True
configMgr.toySeed = 400
# Pruning
# - any overallSys systematic uncertainty if the difference of between the up variation and the nominal and between
# the down variation and the nominal is below a certain (user) given threshold
# - for histoSys types, the situation is more complex:
# - a first check is done if the integral of the up histogram - the integral of the nominal histogram is smaller
# than the integral of the nominal histogram and the same for the down histogram
# - then a second check is done if the shape of the up, down and nominal histograms is very similar Only when both
# conditions are fulfilled the systematics will be removed.
# default is False, so the pruning is normally not enabled
configMgr.prun = True
# The threshold to decide if an uncertainty is small or not is set by configMgr.prunThreshold = 0.005
# where the number gives the fraction of deviation with respect to the nominal histogram below which an uncertainty
# is considered to be small. The default is currently set to 0.01, corresponding to 1 % (This might be very aggressive
# for the one or the other analyses!)
configMgr.prunThreshold = 0.005
# method 1: a chi2 test (this is still a bit experimental, so watch out if this is working or not)
# method 2: checking for every bin of the histograms that the difference between up variation and nominal and down (default)
configMgr.prunMethod = 2
# variation and nominal is below a certain threshold.
# Smoothing: HistFitter does not provide any smoothing tools.
# More Details: https://twiki.cern.ch/twiki/bin/viewauth/AtlasProtected/HistFitterAdvancedTutorial#Pruning_in_HistFitter
##########################
# Keep SRs also in background fit confuguration
configMgr.keepSignalRegionType = True
configMgr.blindSR = BLIND
# Give the analysis a name
configMgr.analysisName = "bbtautau" + "X" + mass
configMgr.histCacheFile = "data/" + configMgr.analysisName + ".root"
configMgr.outputFileName = "results/" + configMgr.analysisName + "_Output.root"
# Define cuts
configMgr.cutsDict["SR"] = "1."
# Define weights
configMgr.weights = "1."
# Define samples
list_samples = []
yields_mass = yields[mass]
for process, yields_process in yields_mass.items():
if process == 'data' or signal_prefix in process: continue
# print("-> {} / Colour: {}".format(process, color_dict[process]))
bkg = Sample(str(process), color_dict[process])
bkg.setStatConfig(stat_config)
# OLD: add lumi uncertainty (bkg/sig correlated, not for data-driven fakes)
# NOW: add lumi by hand
bkg.setNormByTheory(False)
noms = yields_process["nEvents"]
errors = yields_process["nEventsErr"] if use_mcstat else [0.0]
# print(" nEvents (StatError): {} ({})".format(noms, errors))
bkg.buildHisto(noms, "SR", my_disc, 0.5)
bkg.buildStatErrors(errors, "SR", my_disc)
if not stat_only and not no_syst:
if process == 'fakes':
key_here = "ATLAS_FF_1BTAG_SIDEBAND_Syst_hadhad"
if not impact_check_continue(dict_syst_check, key_here):
bkg.addSystematic(
Systematic(key_here, configMgr.weights, 1.50, 0.50,
"user", syst_type))
else:
key_here = "ATLAS_Lumi_Run2_hadhad"
if not impact_check_continue(dict_syst_check, key_here):
bkg.addSystematic(
Systematic(key_here, configMgr.weights, 1.017, 0.983,
"user", syst_type))
for key, values in yields_process.items():
if 'ATLAS' not in key: continue
if impact_check_continue(dict_syst_check, key): continue
# this should not be applied on the Sherpa
if process == 'Zhf' and key == 'ATLAS_DiTauSF_ZMODEL_hadhad':
continue
if process == 'Zlf' and key == 'ATLAS_DiTauSF_ZMODEL_hadhad':
continue
ups = values[0]
downs = values[1]
systUpRatio = [
u / n if n != 0. else float(1.) for u, n in zip(ups, noms)
]
systDoRatio = [
d / n if n != 0. else float(1.)
for d, n in zip(downs, noms)
]
bkg.addSystematic(
Systematic(str(key), configMgr.weights, systUpRatio,
systDoRatio, "user", syst_type))
list_samples.append(bkg)
# FIXME: This is unusual!
top = Sample('top', kOrange)
top.setStatConfig(False) # No stat error
top.setNormByTheory(False) # consider lumi for it
top.buildHisto([0.00001], "SR", my_disc, 0.5) # small enough
# HistFitter can accept such large up ratio
# Systematic(name, weight, ratio_up, ratio_down, syst_type, syst_fistfactory_type)
if not stat_only and not no_syst:
key_here = 'ATLAS_TTBAR_YIELD_UPPER_hadhad'
if not impact_check_continue(dict_syst_check, key_here):
top.addSystematic(
Systematic(key_here, configMgr.weights, unc_ttbar[mass], 0.9,
"user", syst_type))
list_samples.append(top)
sigSample = Sample("Sig", kRed)
sigSample.setNormFactor("mu_Sig", 1., 0., 100.)
#sigSample.setStatConfig(stat_config)
sigSample.setStatConfig(False)
sigSample.setNormByTheory(False)
noms = yields_mass[signal_prefix + mass]["nEvents"]
errors = yields_mass[signal_prefix +
mass]["nEventsErr"] if use_mcstat else [0.0]
sigSample.buildHisto([n * MY_SIGNAL_NORM * 1e-3 for n in noms], "SR",
my_disc, 0.5)
#sigSample.buildStatErrors(errors, "SR", my_disc)
for key, values in yields_mass[signal_prefix + mass].items():
if 'ATLAS' not in key: continue
if impact_check_continue(dict_syst_check, key):
continue
ups = values[0]
downs = values[1]
systUpRatio = [
u / n if n != 0. else float(1.) for u, n in zip(ups, noms)
]
systDoRatio = [
d / n if n != 0. else float(1.) for d, n in zip(downs, noms)
]
if not stat_only and not no_syst:
sigSample.addSystematic(
Systematic(str(key), configMgr.weights, systUpRatio,
systDoRatio, "user", syst_type))
if not stat_only and not no_syst:
key_here = "ATLAS_SigAccUnc_hadhad"
if not impact_check_continue(dict_syst_check, key_here):
sigSample.addSystematic(
Systematic(key_here, configMgr.weights,
[1 + unc_sig_acc[mass] for i in range(my_nbins)],
[1 - unc_sig_acc[mass]
for i in range(my_nbins)], "user", syst_type))
key_here = "ATLAS_Lumi_Run2_hadhad"
if not impact_check_continue(dict_syst_check, key_here):
sigSample.addSystematic(
Systematic(key_here, configMgr.weights, 1.017, 0.983, "user",
syst_type))
list_samples.append(sigSample)
# Set observed and expected number of events in counting experiment
n_SPlusB = yields_mass[signal_prefix +
mass]["nEvents"][0] + sum_of_bkg(yields_mass)[0]
n_BOnly = sum_of_bkg(yields_mass)[0]
if BLIND:
# configMgr.useAsimovSet = True # Use the Asimov dataset
# configMgr.generateAsimovDataForObserved = True # Generate Asimov data as obsData for UL
# configMgr.useSignalInBlindedData = False
ndata = sum_of_bkg(yields_mass)
else:
try:
ndata = yields_mass["data"]["nEvents"]
except:
ndata = [0. for _ in range(my_nbins)]
lumiError = 0.017 # Relative luminosity uncertainty
dataSample = Sample("Data", kBlack)
dataSample.setData()
dataSample.buildHisto(ndata, "SR", my_disc, 0.5)
list_samples.append(dataSample)
# Define top-level
ana = configMgr.addFitConfig("SPlusB")
ana.addSamples(list_samples)
ana.setSignalSample(sigSample)
# Define measurement
meas = ana.addMeasurement(name="NormalMeasurement",
lumi=1.0,
lumiErr=lumiError / 100000.)
# make it very small so that pruned
# we use the one added by hand
meas.addPOI("mu_Sig")
#meas.statErrorType = "Poisson"
# Fix the luminosity in HistFactory to constant
meas.addParamSetting("Lumi", True, 1)
# Add the channel
chan = ana.addChannel(my_disc, ["SR"], my_nbins, my_xmin, my_xmax)
chan.blind = BLIND
#chan.statErrorType = "Poisson"
ana.addSignalChannels([chan])
# These lines are needed for the user analysis to run
# Make sure file is re-made when executing HistFactory
if configMgr.executeHistFactory:
if os.path.isfile("data/%s.root" % configMgr.analysisName):
os.remove("data/%s.root" % configMgr.analysisName)
