def prepare(baseDir,
particle,
probe,
resonance,
era,
config,
num,
denom,
variableLabels,
skipPlots=False,
cutAndCount=False):
subEra = era.split('_')[0] # data subera is beginning of era
lumi = registry.luminosity(particle, probe, resonance, era, subEra)
hists = {}
effName = get_eff_name(num, denom)
extEffName = get_extended_eff_name(num, denom, variableLabels)
binning = config.binning()
dataSubEra, mcSubEra = get_data_mc_sub_eras(resonance, era)
systList = config.get('systematics', {
x: {
'fitTypes': [],
'shiftTypes': []
}
for x in ['SF', 'dataEff', 'mcEff']
})
def get_variable_name_pretty(variableLabel):
variables = config.variables()
return variables.get(variableLabel, {}).get('pretty', variableLabel)
# create output histograms
nVars = len(variableLabels)
if nVars == 1:
THX = ROOT.TH1F
elif nVars == 2:
THX = ROOT.TH2F
elif nVars == 3:
THX = ROOT.TH3F
else:
raise NotImplementedError(
'More than 3 dimensions are not supported for scale factors')
hargs = [extEffName, extEffName]
for variableLabel in variableLabels:
hargs += [
len(binning[variableLabel]) - 1,
array('d', binning[variableLabel])
]
hist = THX(*hargs)
axes = [hist.GetXaxis(), hist.GetYaxis(), hist.GetZaxis()]
for vi, variableLabel in enumerate(variableLabels):
axes[vi].SetTitle(get_variable_name_pretty(variableLabel))
if nVars == 1:
hist.GetYaxis().SetTitle('Scalefactor')
if nVars == 2:
hist.SetOption('colz')
hist.GetZaxis().SetTitle('Scalefactor')
hist_stat = hist.Clone(extEffName + '_stat')
hist_syst = hist.Clone(extEffName + '_syst')
histList_syst = {
'combined_syst': hist.Clone(extEffName + '_combined_syst'),
}
histList_syst['combined_syst'].GetZaxis().SetTitle('Uncertainty')
hist_dataEff = hist.Clone(extEffName + '_efficiencyData')
if nVars == 1:
hist_dataEff.GetYaxis().SetTitle('Efficiency')
if nVars == 2:
hist_dataEff.GetZaxis().SetTitle('Efficiency')
hist_dataEff_stat = hist_dataEff.Clone(extEffName + '_efficiencyData_stat')
hist_dataEff_syst = hist_dataEff.Clone(extEffName + '_efficiencyData_syst')
histList_dataEff_syst = {
'combined_syst':
hist_dataEff.Clone(extEffName + '_efficiencyData_combined_syst'),
}
histList_dataEff_syst['combined_syst'].GetZaxis().SetTitle('Uncertainty')
hist_mcEff = hist_dataEff.Clone(extEffName + '_efficiencyMC')
hist_mcEff_stat = hist_dataEff.Clone(extEffName + '_efficiencyMC_stat')
hist_mcEff_syst = hist_dataEff.Clone(extEffName + '_efficiencyMC_syst')
histList_mcEff_syst = {
'combined_syst':
hist_dataEff.Clone(extEffName + '_efficiencyMC_combined_syst'),
}
histList_mcEff_syst['combined_syst'].GetZaxis().SetTitle('Uncertainty')
# the individual systematics
for iSyst in itertools.chain(systList['SF']['fitTypes'],
systList['SF']['shiftTypes']):
histList_syst[iSyst] = hist.Clone(extEffName + '_' + iSyst)
histList_syst[iSyst + '_syst'] = hist.Clone(extEffName + '_' + iSyst +
'_syst')
histList_syst[iSyst + '_syst'].GetZaxis().SetTitle('Uncertainty')
for iSyst in itertools.chain(systList['dataEff']['fitTypes'],
systList['dataEff']['shiftTypes']):
histList_dataEff_syst[iSyst] = hist_dataEff.Clone(extEffName + '_' +
iSyst)
histList_dataEff_syst[iSyst +
'_syst'] = hist_dataEff.Clone(extEffName + '_' +
iSyst + '_syst')
histList_dataEff_syst[iSyst +
'_syst'].GetZaxis().SetTitle('Uncertainty')
for iSyst in itertools.chain(systList['mcEff']['fitTypes'],
systList['mcEff']['shiftTypes']):
histList_mcEff_syst[iSyst] = hist_mcEff.Clone(extEffName + '_' + iSyst)
histList_mcEff_syst[iSyst +
'_syst'] = hist_mcEff.Clone(extEffName + '_' +
iSyst + '_syst')
histList_mcEff_syst[iSyst + '_syst'].GetZaxis().SetTitle('Uncertainty')
varName = get_variables_name(variableLabels)
# iterate through the bin indices
# this does nested for loops of the N-D binning (e.g. pt, eta)
# binning starts at 1 (0 is underflow), same as ROOT
indices = [
list(range(1, len(binning[variableLabel])))
for variableLabel in variableLabels
]
output = {effName: {varName: {}}}
for index in itertools.product(*indices):
binName = get_full_name(num, denom, variableLabels, index)
subVarKeys = [
'{}:[{},{}]'.format(variableLabels[i],
binning[variableLabels[i]][ind - 1],
binning[variableLabels[i]][ind])
for i, ind in enumerate(index)
]
_out = output[effName][varName]
# add binning definitions
_out['binning'] = [{
'variable': vl,
'binning': binning[vl].tolist(),
} for vl in variableLabels]
for subVarKey in subVarKeys:
if subVarKey not in _out:
_out[subVarKey] = {}
_out = _out[subVarKey]
# the fitted distributions
fitType = 'Nominal'
dataFNameFit = os.path.join(baseDir, 'fits_data', particle, probe,
resonance, era, fitType, effName,
binName + '.root')
dataFNameCNC = os.path.join(baseDir, 'flat', particle, probe,
resonance, era, dataSubEra, 'Nominal',
extEffName + '.root')
mcFNameCNC = os.path.join(baseDir, 'flat', particle, probe, resonance,
era, mcSubEra, 'Nominal',
extEffName + '.root')
if cutAndCount:
sf, sf_stat, dataEff, dataStat, mcEff, mcStat = getSF_cutAndCount(
binName, dataFNameCNC, mcFNameCNC)
else:
sf, sf_stat, dataEff, dataStat, mcEff, mcStat = getSF(
binName, dataFNameFit)
fitTypes = set(systList['SF']['fitTypes'] +
systList['dataEff']['fitTypes'] +
systList['mcEff']['fitTypes'])
shiftTypes = set(systList['SF']['shiftTypes'] +
systList['dataEff']['shiftTypes'] +
systList['mcEff']['shiftTypes'])
if cutAndCount:
sf_syst = getSyst_cutAndCount(binName, dataFNameCNC, mcFNameCNC,
fitTypes, shiftTypes)
else:
sf_syst = getSyst(binName, dataFNameFit, fitTypes, shiftTypes)
combined_syst = {}
for kind in ['SF', 'dataEff', 'mcEff']:
combined_syst[kind] = 0
errKey = 'err'
if kind == 'dataEff':
errKey = 'dataErr'
if kind == 'mcEff':
errKey = 'mcErr'
for t in itertools.chain(systList[kind]['fitTypes'],
systList[kind]['shiftTypes']):
combined_syst[kind] += sf_syst[t][errKey]**2
combined_syst[kind] = combined_syst[kind]**0.5
sf_err = (sf_stat**2 + combined_syst['SF']**2)**0.5
dataErr = (dataStat**2 + combined_syst['dataEff']**2)**0.5
mcErr = (mcStat**2 + combined_syst['mcEff']**2)**0.5
_out['value'] = sf
_out['stat'] = sf_stat
_out['syst'] = combined_syst['SF']
for s in itertools.chain(systList['SF']['fitTypes'],
systList['SF']['shiftTypes']):
_out[s] = sf_syst[s]['err']
def set_bin(hist, index, val, err):
index = list(index)
val_args = index + [val]
err_args = index + [err]
hist.SetBinContent(*val_args)
if err >= 0:
hist.SetBinError(*err_args)
set_bin(hist, index, sf, sf_err)
set_bin(hist_stat, index, sf, sf_stat)
set_bin(hist_syst, index, sf, combined_syst['SF'])
set_bin(histList_syst['combined_syst'], index, combined_syst['SF'], -1)
set_bin(hist_dataEff, index, dataEff, dataErr)
set_bin(hist_dataEff_stat, index, dataEff, dataStat)
set_bin(hist_dataEff_syst, index, dataEff, combined_syst['dataEff'])
set_bin(histList_dataEff_syst['combined_syst'], index,
combined_syst['dataEff'], -1)
set_bin(hist_mcEff, index, mcEff, mcErr)
set_bin(hist_mcEff_stat, index, mcEff, mcStat)
set_bin(hist_mcEff_syst, index, mcEff, combined_syst['mcEff'])
set_bin(histList_mcEff_syst['combined_syst'], index,
combined_syst['mcEff'], -1)
for iKey in sf_syst.keys():
if iKey in histList_syst:
set_bin(histList_syst[iKey], index, sf_syst[iKey]['sf'],
sf_syst[iKey]['err'])
set_bin(histList_syst[iKey + '_syst'], index,
sf_syst[iKey]['err'], -1)
if iKey in histList_dataEff_syst:
set_bin(histList_dataEff_syst[iKey], index,
sf_syst[iKey]['dataEff'], sf_syst[iKey]['dataErr'])
set_bin(histList_dataEff_syst[iKey + '_syst'], index,
sf_syst[iKey]['dataErr'], -1)
if iKey in histList_mcEff_syst:
set_bin(histList_mcEff_syst[iKey], index,
sf_syst[iKey]['mcEff'], sf_syst[iKey]['mcErr'])
set_bin(histList_mcEff_syst[iKey + '_syst'], index,
sf_syst[iKey]['mcErr'], -1)
hists[extEffName] = hist
hists[extEffName + '_stat'] = hist_stat
hists[extEffName + '_syst'] = hist_syst
hists[extEffName + '_efficiencyData'] = hist_dataEff
hists[extEffName + '_efficiencyData_stat'] = hist_dataEff_stat
hists[extEffName + '_efficiencyData_syst'] = hist_dataEff_syst
hists[extEffName + '_efficiencyMC'] = hist_mcEff
hists[extEffName + '_efficiencyMC_stat'] = hist_mcEff_stat
hists[extEffName + '_efficiencyMC_syst'] = hist_mcEff_syst
for iKey in histList_syst.keys():
hname = extEffName + '_' + iKey
hists[hname] = histList_syst[iKey]
for iKey in histList_dataEff_syst.keys():
hname = extEffName + '_efficiencyData_' + iKey
hists[hname] = histList_dataEff_syst[iKey]
for iKey in histList_mcEff_syst.keys():
hname = extEffName + '_efficiencyMC_' + iKey
hists[hname] = histList_mcEff_syst[iKey]
# save the efficiency
plotDir = os.path.join(baseDir, 'plots', particle, probe, resonance, era,
effName, 'efficiency')
os.makedirs(plotDir, exist_ok=True)
effDir = os.path.join(baseDir, 'efficiencies', particle, probe, resonance,
era, effName)
os.makedirs(effDir, exist_ok=True)
effPath = os.path.join(effDir, extEffName)
# JSON format
with open('{}.json'.format(effPath), 'w') as f:
f.write(json.dumps(output, indent=4, sort_keys=True))
# ROOT histogram format
tfile = ROOT.TFile.Open('{}.root'.format(effPath), 'recreate')
for h in sorted(hists):
hists[h].Write(h)
if nVars == 2 and not skipPlots:
cName = 'c' + h
canvas = ROOT.TCanvas(cName, cName, 1000, 800)
ROOT.gStyle.SetPaintTextFormat("5.3f")
canvas.SetRightMargin(0.24)
hists[h].Draw('colz text')
plotPath = os.path.join(plotDir, h)
canvas.Modified()
canvas.Update()
CMS_lumi.cmsText = 'CMS'
CMS_lumi.writeExtraText = True
CMS_lumi.extraText = 'Preliminary'
CMS_lumi.lumi_13TeV = "%0.1f fb^{-1}" % (lumi)
CMS_lumi.CMS_lumi(canvas, 4, 0)
canvas.Print('{}.png'.format(plotPath))
canvas.Print('{}.pdf'.format(plotPath))
canvas.Print('{}.root'.format(plotPath))
tfile.Close()
if skipPlots:
return
# gets a graph projection of an ND histogram for a given axis
# with axis index (ie x,y,z = 0,1,2) and other dimensions ind
def get_graph(hist, axis, axis_ind, *ind):
ind = list(ind)
ni = axis.GetNbins()
xvals = [axis.GetBinCenter(i + 1) for i in range(ni)]
xvals_errLow = [
xvals[i] - axis.GetBinLowEdge(i + 1) for i in range(ni)
]
xvals_errHigh = [
axis.GetBinUpEdge(i + 1) - xvals[i] for i in range(ni)
]
yvals = [
hist.GetBinContent(*ind[:axis_ind] + [i + 1] + ind[axis_ind:])
for i in range(ni)
]
yvals_err = [
hist.GetBinError(*ind[:axis_ind] + [i + 1] + ind[axis_ind:])
for i in range(ni)
]
graph = ROOT.TGraphAsymmErrors(
ni,
array('d', xvals),
array('d', yvals),
array('d', xvals_errLow),
array('d', xvals_errHigh),
array('d', yvals_err),
array('d', yvals_err),
)
return graph
# plot the efficiencies
# some default colors for plots
colors = [
ROOT.kBlack, ROOT.kBlue, ROOT.kRed, ROOT.kGreen + 2, ROOT.kMagenta + 1,
ROOT.kOrange + 1, ROOT.kTeal - 1, ROOT.kRed - 3, ROOT.kCyan + 2
]
def plot_1d_eff(savename,
graphs,
labels=['Data', 'Simulation'],
colors=colors,
xlabel='',
ylabel='Efficiency',
xRange=[],
additional_text=[]):
ng = len(graphs)
mg = ROOT.TMultiGraph()
for gi in range(ng):
graphs[gi].SetLineColor(colors[gi])
graphs[gi].SetMarkerColor(colors[gi])
mg.Add(graphs[gi])
canvas = ROOT.TCanvas(savename, savename, 800, 800)
mg.Draw('AP0')
mg.GetXaxis().SetTitle(xlabel)
if xRange:
mg.GetXaxis().SetRangeUser(*xRange)
mg.GetYaxis().SetTitle(ylabel)
mg.GetYaxis().SetRangeUser(0.8, 1.10)
legend = ROOT.TLegend(0.5, 0.70, 0.92, 0.92)
legend.SetTextFont(42)
legend.SetBorderSize(0)
legend.SetFillColor(0)
for gi in range(ng):
legend.AddEntry(graphs[gi], labels[gi], 'l')
legend.SetHeader('{} / {}'.format(num, denom))
legend.Draw()
if additional_text:
nother = len(additional_text)
dims = [0.18, 0.84 - nother * 0.04 - 0.02, 0.35, 0.84]
text = ROOT.TPaveText(*dims + ['NB NDC'])
text.SetTextFont(42)
text.SetBorderSize(0)
text.SetFillColor(0)
text.SetTextAlign(11)
text.SetTextSize(0.03)
for rtext in additional_text:
text.AddText(rtext)
text.Draw()
CMS_lumi.cmsText = 'CMS'
CMS_lumi.writeExtraText = True
CMS_lumi.extraText = 'Preliminary'
CMS_lumi.lumi_13TeV = "%0.1f fb^{-1}" % (lumi)
CMS_lumi.CMS_lumi(canvas, 4, 11)
canvas.Print('{}.png'.format(savename))
canvas.Print('{}.pdf'.format(savename))
canvas.Print('{}.root'.format(savename))
# enumerate over the axis/variable to plot
axes = [
hists[extEffName].GetXaxis(), hists[extEffName].GetYaxis(),
hists[extEffName].GetZaxis()
]
for vi, variableLabel in enumerate(variableLabels):
# iterate over the other axis indices
otherVariableLabels = [
ovl for ovl in variableLabels if ovl != variableLabel
]
otherVariableIndices = [
ovi for ovi, ovl in enumerate(variableLabels)
if ovl != variableLabel
]
indices = [
list(range(1, len(binning[vl]))) for vl in otherVariableLabels
]
if indices:
for index in itertools.product(*indices):
graph_data = get_graph(hists[extEffName + '_efficiencyData'],
axes[vi], vi, *index)
graph_mc = get_graph(hists[extEffName + '_efficiencyMC'],
axes[vi], vi, *index)
xlabel = get_variable_name_pretty(variableLabel)
ylabel = 'Efficiency'
xRange = [
axes[vi].GetBinLowEdge(1),
axes[vi].GetBinUpEdge(axes[vi].GetNbins())
]
additional_text = []
for novi, (ovi, ovl) in enumerate(
zip(otherVariableIndices, otherVariableLabels)):
xlow = axes[ovi].GetBinLowEdge(index[novi])
xhigh = axes[ovi].GetBinUpEdge(index[novi])
rtext = '{} < {} < {}'.format(
xlow, get_variable_name_pretty(ovl), xhigh)
additional_text += [rtext]
plotDir = os.path.join(baseDir, 'plots', particle, probe,
resonance, era, effName, 'efficiency')
os.makedirs(plotDir, exist_ok=True)
otherVariableLabel = get_bin_name(otherVariableLabels, index)
plotName = '{}_{}_vs_{}'.format(effName, otherVariableLabel,
variableLabel)
plotPath = os.path.join(plotDir, plotName)
plot_1d_eff(plotPath, [graph_data, graph_mc],
xlabel=xlabel,
ylabel=ylabel,
xRange=xRange,
additional_text=additional_text)
# dataEfficiency systs
graphs = [
get_graph(hists[extEffName + '_efficiencyData'], axes[vi],
vi, *index)
]
labels = ['Nominal']
for iSyst in itertools.chain(
systList['dataEff']['fitTypes'],
systList['dataEff']['shiftTypes']):
graphs += [
get_graph(
hists[extEffName + '_efficiencyData_' + iSyst],
axes[vi], vi, *index)
]
labels += [iSyst]
plotName = '{}_{}_vs_{}_efficiencyData_syst'.format(
effName,
otherVariableLabel,
variableLabel,
)
plotPath = os.path.join(plotDir, plotName)
plot_1d_eff(plotPath,
graphs,
labels=labels,
xlabel=xlabel,
ylabel=ylabel,
xRange=xRange,
additional_text=additional_text)
# mcEfficiency systs
graphs = [
get_graph(hists[extEffName + '_efficiencyMC'], axes[vi],
vi, *index)
]
labels = ['Nominal']
for iSyst in itertools.chain(systList['mcEff']['fitTypes'],
systList['mcEff']['shiftTypes']):
graphs += [
get_graph(hists[extEffName + '_efficiencyMC_' + iSyst],
axes[vi], vi, *index)
]
labels += [iSyst]
plotName = '{}_{}_vs_{}_efficiencyMC_syst'.format(
effName,
otherVariableLabel,
variableLabel,
)
plotPath = os.path.join(plotDir, plotName)
plot_1d_eff(plotPath,
graphs,
labels=labels,
xlabel=xlabel,
ylabel=ylabel,
xRange=xRange,
additional_text=additional_text)
# if no indices, easier, just itself
else:
graph_data = get_graph(hists[extEffName + '_efficiencyData'],
axes[vi], vi)
graph_mc = get_graph(hists[extEffName + '_efficiencyMC'], axes[vi],
vi)
xlabel = get_variable_name_pretty(variableLabel)
ylabel = 'Efficiency'
xRange = [
axes[0].GetBinLowEdge(1),
axes[0].GetBinUpEdge(axes[0].GetNbins())
]
plotDir = os.path.join(baseDir, 'plots', particle, probe,
resonance, era, effName, 'efficiency')
os.makedirs(plotDir, exist_ok=True)
plotName = '{}_vs_{}'.format(effName, variableLabel)
plotPath = os.path.join(plotDir, plotName)
plot_1d_eff(plotPath, [graph_data, graph_mc],
xlabel=xlabel,
ylabel=ylabel,
xRange=xRange)
def prepare(baseDir, particle, probe, resonance, era, config, num, denom,
variableLabels):
hists = {}
effName = get_eff_name(num, denom)
extEffName = get_extended_eff_name(num, denom, variableLabels)
binning = config.binning()
dataSubEra, mcSubEra = get_data_mc_sub_eras(resonance, era)
systList = config.get('systematics', {
x: {
'fitTypes': [],
'shittTypes': []
}
for x in ['SF', 'dataEff', 'mcEff']
})
def get_variable_name_pretty(variableLabel):
variables = config.variables()
return variables.get(variableLabel, {}).get('pretty', variableLabel)
# create output histograms
nVars = len(variableLabels)
if nVars == 1:
THX = ROOT.TH1F
elif nVars == 2:
THX = ROOT.TH2F
elif nVars == 3:
THX = ROOT.TH3F
else:
raise NotImplementedError(
'More than 3 dimensions are not supported for scale factors')
hargs = [extEffName, extEffName]
for variableLabel in variableLabels:
hargs += [
len(binning[variableLabel]) - 1,
array('d', binning[variableLabel])
]
hist = THX(*hargs)
axes = [hist.GetXaxis(), hist.GetYaxis(), hist.GetZaxis()]
for vi, variableLabel in enumerate(variableLabels):
axes[vi].SetTitle(get_variable_name_pretty(variableLabel))
if nVars == 1:
hist.GetYaxis().SetTitle('Scalefactor')
if nVars == 2:
hist.SetOption('colz')
hist.GetZaxis().SetTitle('Scalefactor')
hist_stat = hist.Clone(extEffName + '_stat')
hist_syst = hist.Clone(extEffName + '_syst')
histList_syst = {'combined': hist.Clone(effName + '_combinedSyst')}
hist_dataEff = hist.Clone(extEffName + '_efficiencyData')
if nVars == 1:
hist_dataEff.GetYaxis().SetTitle('Efficiency')
if nVars == 2:
hist_dataEff.GetZaxis().SetTitle('Efficiency')
hist_dataEff_stat = hist_dataEff.Clone(extEffName + '_efficiencyData_stat')
hist_dataEff_syst = hist_dataEff.Clone(extEffName + '_efficiencyData_syst')
histList_dataEff_syst = {
'combined':
hist_dataEff.Clone(effName + '_efficiencyData_combinedSyst')
}
hist_mcEff = hist_dataEff.Clone(extEffName + '_efficiencyMC')
hist_mcEff_stat = hist_dataEff.Clone(extEffName + '_efficiencyMC_stat')
hist_mcEff_syst = hist_dataEff.Clone(extEffName + '_efficiencyMC_syst')
histList_mcEff_syst = {
'combined': hist_dataEff.Clone(effName + '_efficiencyMC_combinedSyst')
}
for iSyst in itertools.chain(systList['SF']['fitTypes'],
systList['SF']['shiftTypes']):
histList_syst.update({iSyst: hist.Clone(effName + '_' + iSyst)})
for iSyst in itertools.chain(systList['dataEff']['fitTypes'],
systList['dataEff']['shiftTypes']):
histList_dataEff_syst.update(
{iSyst: hist.Clone(effName + '_' + iSyst)})
for iSyst in itertools.chain(systList['mcEff']['fitTypes'],
systList['mcEff']['shiftTypes']):
histList_mcEff_syst.update({iSyst: hist.Clone(effName + '_' + iSyst)})
varName = get_variables_name(variableLabels)
# iterate through the bin indices
# this does nested for loops of the N-D binning (e.g. pt, eta)
# binning starts at 1 (0 is underflow), same as ROOT
indices = [
list(range(1, len(binning[variableLabel])))
for variableLabel in variableLabels
]
output = {effName: {varName: {}}}
for index in itertools.product(*indices):
binName = get_full_name(num, denom, variableLabels, index)
subVarKeys = [
'{}:[{},{}]'.format(variableLabels[i],
binning[variableLabels[i]][ind - 1],
binning[variableLabels[i]][ind])
for i, ind in enumerate(index)
]
_out = output[effName][varName]
# add binning definitions
_out['binning'] = [{
'variable': vl,
'binning': binning[vl].tolist(),
} for vl in variableLabels]
for subVarKey in subVarKeys:
if subVarKey not in _out:
_out[subVarKey] = {}
_out = _out[subVarKey]
# the fitted distributions
fitType = 'Nominal'
dataFNameFit = os.path.join(baseDir, 'fits_data', particle, probe,
resonance, era, fitType, effName,
binName + '.root')
sf, sf_stat, dataEff, dataStat, mcEff, mcStat = getSF(
binName, dataFNameFit)
sf_syst = getSyst(binName, dataFNameFit, dataEff, mcEff,
systList['SF']['fitTypes'],
systList['SF']['shiftTypes'])
dataSyst = getSyst(binName, dataFNameFit, dataEff, mcEff,
systList['dataEff']['fitTypes'],
systList['dataEff']['shiftTypes'])
mcSyst = getSyst(binName, dataFNameFit, dataEff, mcEff,
systList['mcEff']['fitTypes'],
systList['mcEff']['shiftTypes'])
sf_err = (sf_stat**2 + sf_syst['combined']**2)**0.5
dataErr = (dataStat**2 + dataSyst['combined']**2)**0.5
mcErr = (mcStat**2 + mcSyst['combined']**2)**0.5
_out['value'] = sf
_out['stat'] = sf_stat
_out['syst'] = sf_syst['combined']
for s in itertools.chain(systList['SF']['fitTypes'],
systList['SF']['shiftTypes']):
_out[s] = sf_syst[s]
def set_bin(hist, index, val, err):
index = list(index)
val_args = index + [val]
err_args = index + [err]
hist.SetBinContent(*val_args)
if err >= 0:
hist.SetBinError(*err_args)
set_bin(hist, index, sf, sf_err)
set_bin(hist_stat, index, sf, sf_stat)
set_bin(hist_syst, index, sf, sf_syst['combined'])
for iKey in sf_syst.keys():
set_bin(histList_syst[iKey], index, sf_syst[iKey], -1)
set_bin(hist_dataEff, index, dataEff, dataErr)
set_bin(hist_dataEff_stat, index, dataEff, dataStat)
set_bin(hist_dataEff_syst, index, dataEff, dataSyst['combined'])
for iKey in dataSyst.keys():
set_bin(histList_dataEff_syst[iKey], index, dataSyst[iKey], -1)
set_bin(hist_mcEff, index, mcEff, mcErr)
set_bin(hist_mcEff_stat, index, mcEff, mcStat)
set_bin(hist_mcEff_syst, index, mcEff, mcSyst['combined'])
for iKey in mcSyst.keys():
set_bin(histList_mcEff_syst[iKey], index, mcSyst[iKey], -1)
hists[extEffName] = hist
hists[extEffName + '_stat'] = hist_stat
hists[extEffName + '_syst'] = hist_syst
hists[extEffName + '_efficiencyData'] = hist_dataEff
hists[extEffName + '_efficiencyData_stat'] = hist_dataEff_stat
hists[extEffName + '_efficiencyData_syst'] = hist_dataEff_syst
hists[extEffName + '_efficiencyMC'] = hist_mcEff
hists[extEffName + '_efficiencyMC_stat'] = hist_mcEff_stat
hists[extEffName + '_efficiencyMC_syst'] = hist_mcEff_syst
for iKey in histList_syst.keys():
hists[effName + '_' + iKey] = histList_syst[iKey]
for iKey in histList_dataEff_syst.keys():
hists[effName + '_efficiencyData_' +
iKey] = histList_dataEff_syst[iKey]
for iKey in histList_mcEff_syst.keys():
hists[effName + '_efficiencyMC_' + iKey] = histList_mcEff_syst[iKey]
# save the efficiency
plotDir = os.path.join(baseDir, 'plots', particle, probe, resonance, era,
effName, 'efficiency')
os.makedirs(plotDir, exist_ok=True)
effDir = os.path.join(baseDir, 'efficiencies', particle, probe, resonance,
era, effName)
os.makedirs(effDir, exist_ok=True)
effPath = os.path.join(effDir, extEffName)
# JSON format
with open('{}.json'.format(effPath), 'w') as f:
f.write(json.dumps(output, indent=4, sort_keys=True))
# ROOT histogram format
tfile = ROOT.TFile.Open('{}.root'.format(effPath), 'recreate')
for h in sorted(hists):
hists[h].Write(h)
if nVars == 2:
cName = 'c' + h
canvas = ROOT.TCanvas(cName, cName, 1000, 800)
ROOT.gStyle.SetPaintTextFormat("5.3f")
canvas.SetRightMargin(0.24)
hists[h].Draw('colz text')
plotPath = os.path.join(plotDir, h)
canvas.Modified()
canvas.Update()
canvas.Print('{}.png'.format(plotPath))
canvas.Print('{}.pdf'.format(plotPath))
tfile.Close()
# gets a graph projection of an ND histogram for a given axis
# with axis index (ie x,y,z = 0,1,2) and other dimensions ind
def get_graph(hist, axis, axis_ind, *ind):
ind = list(ind)
ni = axis.GetNbins()
xvals = [axis.GetBinCenter(i + 1) for i in range(ni)]
xvals_errLow = [
xvals[i] - axis.GetBinLowEdge(i + 1) for i in range(ni)
]
xvals_errHigh = [
axis.GetBinUpEdge(i + 1) - xvals[i] for i in range(ni)
]
yvals = [
hist.GetBinContent(*ind[:axis_ind] + [i + 1] + ind[axis_ind:])
for i in range(ni)
]
yvals_err = [
hist.GetBinError(*ind[:axis_ind] + [i + 1] + ind[axis_ind:])
for i in range(ni)
]
graph = ROOT.TGraphAsymmErrors(
ni,
array('d', xvals),
array('d', yvals),
array('d', xvals_errLow),
array('d', xvals_errHigh),
array('d', yvals_err),
array('d', yvals_err),
)
return graph
# plot the efficiencies
# enumerate over the axis/variable to plot
axes = [
hists[extEffName].GetXaxis(), hists[extEffName].GetYaxis(),
hists[extEffName].GetZaxis()
]
for vi, variableLabel in enumerate(variableLabels):
# iterate over the other axis indices
otherVariableLabels = [
ovl for ovl in variableLabels if ovl != variableLabel
]
otherVariableIndices = [
ovi for ovi, ovl in enumerate(variableLabels)
if ovl != variableLabel
]
indices = [
list(range(1, len(binning[vl]))) for vl in otherVariableLabels
]
if indices:
for index in itertools.product(*indices):
graph_data = get_graph(hists[extEffName + '_efficiencyData'],
axes[vi], vi, *index)
graph_data.SetLineColor(ROOT.kBlack)
graph_data.SetMarkerColor(ROOT.kBlack)
graph_mc = get_graph(hists[extEffName + '_efficiencyMC'],
axes[vi], vi, *index)
graph_mc.SetLineColor(ROOT.kBlue)
graph_mc.SetMarkerColor(ROOT.kBlue)
mg = ROOT.TMultiGraph()
mg.Add(graph_data)
mg.Add(graph_mc)
cName = 'c' + extEffName + '_'.join([str(i) for i in index])\
+ variableLabel
canvas = ROOT.TCanvas(cName, cName, 800, 800)
mg.Draw('AP0')
mg.GetXaxis().SetTitle(get_variable_name_pretty(variableLabel))
xRange = [
axes[vi].GetBinLowEdge(1),
axes[vi].GetBinUpEdge(axes[vi].GetNbins())
]
mg.GetXaxis().SetRangeUser(*xRange)
mg.GetYaxis().SetTitle('Efficiency')
mg.GetYaxis().SetRangeUser(0.8, 1.10)
legend = ROOT.TLegend(0.5, 0.70, 0.92, 0.92)
legend.SetTextFont(42)
legend.SetBorderSize(0)
legend.SetFillColor(0)
legend.AddEntry(graph_data, 'Data', 'l')
legend.AddEntry(graph_mc, 'Simulation', 'l')
legend.SetHeader('{} / {}'.format(num, denom))
legend.Draw()
nother = len(indices)
dims = [0.18, 0.84 - nother * 0.04 - 0.02, 0.35, 0.84]
text = ROOT.TPaveText(*dims + ['NB NDC'])
text.SetTextFont(42)
text.SetBorderSize(0)
text.SetFillColor(0)
text.SetTextAlign(11)
text.SetTextSize(0.03)
for novi, (ovi, ovl) in enumerate(
zip(otherVariableIndices, otherVariableLabels)):
xlow = axes[ovi].GetBinLowEdge(index[novi])
xhigh = axes[ovi].GetBinUpEdge(index[novi])
rtext = '{} < {} < {}'.format(
xlow, get_variable_name_pretty(ovl), xhigh)
text.AddText(rtext)
text.Draw()
CMS_lumi.cmsText = 'CMS'
CMS_lumi.writeExtraText = True
CMS_lumi.extraText = 'Preliminary'
CMS_lumi.lumi_13TeV = "%0.1f fb^{-1}" % (41.5)
CMS_lumi.CMS_lumi(canvas, 4, 11)
plotDir = os.path.join(baseDir, 'plots', particle, probe,
resonance, era, effName, 'efficiency')
os.makedirs(plotDir, exist_ok=True)
otherVariableLabel = get_bin_name(otherVariableLabels, index)
plotName = '{}_{}_vs_{}'.format(effName, otherVariableLabel,
variableLabel)
plotPath = os.path.join(plotDir, plotName)
canvas.Print('{}.png'.format(plotPath))
canvas.Print('{}.pdf'.format(plotPath))
# if no indices, easier, just itself
else:
graph_data = get_graph(hists[extEffName + '_efficiencyData'],
axes[vi], vi)
graph_data.SetLineColor(ROOT.kBlack)
graph_data.SetMarkerColor(ROOT.kBlack)
graph_mc = get_graph(hists[extEffName + '_efficiencyMC'], axes[vi],
vi)
graph_mc.SetLineColor(ROOT.kBlue)
graph_mc.SetMarkerColor(ROOT.kBlue)
mg = ROOT.TMultiGraph()
mg.Add(graph_data)
mg.Add(graph_mc)
canvas = ROOT.TCanvas('c' + extEffName, 'c', 800, 800)
mg.Draw('AP0')
mg.GetXaxis().SetTitle(get_variable_name_pretty(variableLabel))
mg.GetYaxis().SetTitle('Efficiency')
mg.GetYaxis().SetRangeUser(0.8, 1.10)
legend = ROOT.TLegend(0.5, 0.70, 0.92, 0.92)
legend.SetTextFont(42)
legend.SetBorderSize(0)
legend.SetFillColor(0)
legend.AddEntry(graph_data, 'Data', 'l')
legend.AddEntry(graph_mc, 'Simulation', 'l')
legend.SetHeader('{} / {}'.format(num, denom))
legend.Draw()
CMS_lumi.cmsText = 'CMS'
CMS_lumi.writeExtraText = True
CMS_lumi.extraText = 'Preliminary'
CMS_lumi.lumi_13TeV = "%0.1f fb^{-1}" % (41.5)
CMS_lumi.CMS_lumi(canvas, 4, 11)
plotDir = os.path.join(baseDir, 'plots', particle, probe,
resonance, era, effName, 'efficiency')
os.makedirs(plotDir, exist_ok=True)
plotName = '{}_vs_{}'.format(effName, variableLabel)
plotPath = os.path.join(plotDir, plotName)
canvas.Print('{}.png'.format(plotPath))
canvas.Print('{}.pdf'.format(plotPath))
def run_conversion(spark, particle, probe, resonance, era, subEra,
config, shift='Nominal', **kwargs):
_numerator = kwargs.pop('numerator', [])
_denominator = kwargs.pop('denominator', [])
_baseDir = kwargs.pop('baseDir', '')
testing = False
print('Running conversion for', resonance, era, subEra, shift)
if useParquet:
fnames = list(registry.parquet(
particle, probe, resonance, era, subEra))
else:
fnames = registry.root(particle, probe, resonance, era, subEra)
fnames = ['root://eoscms.cern.ch/'+f for f in fnames]
jobPath = os.path.join(particle, probe, resonance, era, subEra)
if shift:
jobPath = os.path.join(jobPath, shift)
if testing:
jobPath = os.path.join('testing', jobPath)
else:
jobPath = os.path.join('flat', jobPath)
if _baseDir:
jobPath = os.path.join(_baseDir, jobPath)
os.makedirs(jobPath, exist_ok=True)
doGen = subEra in ['DY_madgraph', 'DY_powheg']
# default numerator/denominator defintions
efficiencies = config.efficiencies()
# get the dataframe
if useParquet:
print('Loading parquet files:', fnames)
if isinstance(fnames, list):
baseDF = spark.read.parquet(*fnames)
else:
baseDF = spark.read.parquet(fnames)
else:
treename = registry.treename(particle, probe, resonance, era, subEra)
baseDF = spark.read.format("root")\
.option('tree', treename)\
.load(fnames)
# create the miniIsoaltion columns
miniIsoDF = get_miniIso_dataframe(baseDF)
# create the definitions columns
definitions = config.definitions()
defDF = miniIsoDF
for d in definitions:
defDF = defDF.withColumn(d, F.expr(definitions[d]))
# select tags
tagsDF = defDF.filter(config.selection())
# build the weights (pileup for MC)
weightedDF = get_weighted_dataframe(
tagsDF, doGen, resonance, era, subEra, shift=shift)
# create the binning structure
fitVariable = config.fitVariable()
binningSet = set([fitVariable])
if doGen:
fitVariableGen = config.fitVariableGen()
binningSet = binningSet.union(set([fitVariableGen]))
binVariables = config.binVariables()
for bvs in binVariables:
binningSet = binningSet.union(set(bvs))
binning = config.binning()
variables = config.variables()
binnedDF = weightedDF
for bName in binningSet:
binnedDF = get_binned_dataframe(
binnedDF, bName+"Bin",
variables[bName]['variable'],
binning[bName])
# build the unrealized yield dataframes
# they are binned in the ID, bin variables, and fit variable
yields = {}
yields_gen = {}
for numLabel, denLabel in efficiencies:
den = binnedDF.filter(denLabel)
for binVars in binVariables:
key = (numLabel, denLabel, tuple(binVars))
yields[key] = den.groupBy(
numLabel, *[b+'Bin' for b in
binVars+[fitVariable]])\
.agg({'weight2': 'sum', 'weight': 'sum'})
if doGen:
yields_gen[key] = den.groupBy(
numLabel, *[b+'Bin' for b in
binVars+[fitVariableGen]])\
.agg({'weight2': 'sum', 'weight': 'sum'})
def get_values(df, mLabel, **binValues):
for k, v in binValues.items():
df = df[df[k] == v]
df = df.set_index(mLabel)
# fill empty bins with 0
# includes underflow and overflow in the ROOT numbering scheme
# (0 is underflow, len(binning)+1 is overflow)
values = pd.Series(np.zeros(len(binning['mass'])+1))
values[df.index] = df['sum(weight)']
values = values.to_numpy()
sumw2 = pd.Series(np.zeros(len(binning['mass'])+1))
if 'sum(weight2)' in df.columns:
sumw2[df.index] = df['sum(weight2)']
else:
sumw2[df.index] = df['sum(weight)'] # no weights provided
sumw2 = sumw2.to_numpy()
return values, sumw2
def get_hist(values, sumw2, edges, overflow=True):
if overflow:
hist = TH1.from_numpy((values[1:-1], edges))
hist[0] = values[0]
hist[-1] = values[-1]
hist._fSumw2 = sumw2
else:
hist = TH1.from_numpy((values, edges))
hist._fSumw2[1:-1] = sumw2
return hist
# realize each of the yield tables
# then produce the histograms and saves them
# this is the first time things are put into memory
for num_den_binVars in yields:
num, den, binVars = num_den_binVars
if _numerator and num not in _numerator:
continue
if _denominator and den not in _denominator:
continue
extended_eff_name = get_extended_eff_name(num, den, binVars)
eff_outname = f'{jobPath}/{extended_eff_name}.root'
hists = {}
print('Processing', eff_outname)
realized = yields[num_den_binVars].toPandas()
for bins in itertools.product(
*[range(1, len(binning[b])) for b in binVars]):
binname = get_full_name(num, den, binVars, bins)
binargs = {b+'Bin': v for b, v in zip(binVars, bins)}
mLabel = fitVariable + 'Bin'
passargs = {num: True}
passargs.update(binargs)
values, sumw2 = get_values(realized, mLabel, **passargs)
edges = binning[fitVariable]
hists[binname+'_Pass'] = get_hist(values, sumw2, edges)
failargs = {num: False}
failargs.update(binargs)
values, sumw2 = get_values(realized, mLabel, **failargs)
edges = binning[fitVariable]
hists[binname+'_Fail'] = get_hist(values, sumw2, edges)
if doGen:
realized = yields_gen[num_den_binVars].toPandas()
for bins in itertools.product(
*[range(1, len(binning[b])) for b in binVars]):
binname = get_full_name(num, den, binVars, bins)
binargs = {b+'Bin': v for b, v in zip(binVars, bins)}
mLabel = fitVariableGen + 'Bin'
passargs = {num: True}
passargs.update(binargs)
values, sumw2 = get_values(realized, mLabel, **passargs)
edges = binning[fitVariableGen]
hists[binname+'_Pass_Gen'] = get_hist(values, sumw2, edges)
failargs = {num: False}
failargs.update(binargs)
values, sumw2 = get_values(realized, mLabel, **failargs)
edges = binning[fitVariableGen]
hists[binname+'_Fail_Gen'] = get_hist(values, sumw2, edges)
with uproot.recreate(eff_outname) as f:
for h, hist in sorted(hists.items()):
f[h] = hist
def build_fit_jobs(particle, probe, resonance, era, config, **kwargs):
_baseDir = kwargs.pop('baseDir', '')
_numerator = kwargs.pop('numerator', [])
_denominator = kwargs.pop('denominator', [])
_fitType = kwargs.pop('fitType', [])
_shiftType = kwargs.pop('shiftType', [])
_sampleType = kwargs.pop('sampleType', [])
_efficiencyBin = kwargs.pop('efficiencyBin', [])
_recover = kwargs.pop('recover', False)
_recoverMode = kwargs.pop('recoverMode', 'simple')
doData = (not _sampleType) or ('data' in _sampleType)
doMC = (not _sampleType) or ('mc' in _sampleType)
dataSubEra, mcSubEra = get_data_mc_sub_eras(resonance, era)
def process(outFName):
if _recover and _recoverMode == 'simple':
return recover_simple(outFName)
return True
jobs = []
# iterate through the efficiencies
efficiencies = config.efficiencies()
binning = config.binning()
for num, denom in efficiencies:
if _numerator and num not in _numerator:
continue
if _denominator and denom not in _denominator:
continue
# iterate through the output binning structure
for variableLabels in config.binVariables():
# iterate through the bin indices
# this does nested for loops of the N-D binning (e.g. pt, eta)
indices = [
list(range(len(binning[variableLabel]) - 1))
for variableLabel in variableLabels
]
for index in itertools.product(*indices):
# binning goes from 1 to N
index = [i + 1 for i in index]
binName = get_full_name(num, denom, variableLabels, index)
extEffName = get_extended_eff_name(num, denom, variableLabels)
effName = get_eff_name(num, denom)
if _efficiencyBin and binName not in _efficiencyBin:
continue
def get_jobs(fitType, shiftType, inType, outType):
_jobs = []
templateFName = os.path.join(_baseDir, 'flat', particle,
probe, resonance, era,
mcSubEra, inType,
extEffName + '.root')
outFName = os.path.join(_baseDir, 'fits_data', particle,
probe, resonance, era, outType,
effName, binName + '.root')
inFName = os.path.join(_baseDir, 'flat', particle, probe,
resonance, era, dataSubEra, inType,
extEffName + '.root')
plotDir = os.path.join(_baseDir, 'plots', particle, probe,
resonance, era, 'fits_data',
outType, effName)
if doData and process(outFName):
_jobs += [(outFName, inFName, binName, templateFName,
plotDir, fitType, 'data', shiftType)]
outFName = os.path.join(_baseDir, 'fits_mc', particle,
probe, resonance, era, outType,
effName, binName + '.root')
inFName = os.path.join(_baseDir, 'flat', particle, probe,
resonance, era, mcSubEra, inType,
extEffName + '.root')
plotDir = os.path.join(_baseDir, 'plots', particle, probe,
resonance, era, 'fits_mc', outType,
effName)
# there is no need to fit MC for templates
# PDF based fits are:
# NominalOld, AltSigOld
if doMC and process(outFName) and\
fitType in ['NominalOld', 'AltSigOld']:
_jobs += [(outFName, inFName, binName, templateFName,
plotDir, fitType, 'mc', shiftType)]
return _jobs
for fitShift in config.fitShifts():
if (_fitType or _shiftType):
if not ((_fitType and fitShift in _fitType) or
(_shiftType and fitShift in _shiftType)):
continue
params = config.fitShift(fitShift)
jobs += get_jobs(params['fitType'], params['shiftType'],
params['inType'], fitShift)
return jobs
