def lheVarHistsForProcess(self, group, processName, chan, varName):
weighthist_name = self.weightHistName(chan, processName)
weightHist = group.FindObject(weighthist_name)
if not weightHist:
raise ValueError("Failed to find %s. Skipping" % weighthist_name)
var = self.theoryVariations[processName][varName]
if not self.isUnrolledFit:
hists, name = HistTools.getLHEWeightHists(weightHist,
var['entries'], "",
varName)
else:
hists, name = HistTools.getTransformed3DLHEHists(
weightHist, HistTools.makeUnrolledHist,
[self.unrolledBinsX, self.unrolledBinsY], var['entries'], "",
varName)
# In this case, should be central, varCentral, varUp, varDown
if len(var['entries']) == 4:
for i in range(hists[0].GetNbinsX() + 1):
central_ratio = hists[0].GetBinContent(i) / hists[
1].GetBinContent(i) if hists[1].GetBinContent(i) > 0 else 1
hists[2].SetBinContent(
i, hists[2].GetBinContent(i) * central_ratio)
hists[3].SetBinContent(
i, hists[3].GetBinContent(i) * central_ratio)
hists = [hists[2], hists[3]]
hists[0] = HistTools.rebinHist(hists[0], name, self.rebin)
hists[1] = HistTools.rebinHist(hists[1], name.replace("Up", "Down"),
self.rebin)
return hists
def scaleHistsForProcess(self, group, processName, chan, expandedTheory, append=""):
weightHist = group.FindObject(self.weightHistName(chan, processName, append))
if not weightHist:
raise ValueError("Failed to find %s. Skipping" % self.weightHistName(chan, processName, append))
if 'scale' not in self.theoryVariations[processName]:
return []
scaleVars = self.theoryVariations[processName]['scale']
scaleHists = HistTools.getScaleHists(weightHist, processName, self.rebin,
entries=scaleVars['entries'],
exclude=scaleVars['exclude'],
central=scaleVars['central']) if not self.isUnrolledFit else \
HistTools.getTransformed3DScaleHists(weightHist, HistTools.makeUnrolledHist,
[self.unrolledBinsX, self.unrolledBinsY], processName,
entries=scaleVars['entries'],
exclude=scaleVars['exclude'])
if expandedTheory:
name = processName if not self.correlateScaleUnc else ""
expandedScaleHists = HistTools.getExpandedScaleHists(weightHist, name, self.rebin,
entries=scaleVars['entries'],
pairs=scaleVars['groups'],
) if not self.isUnrolledFit else \
HistTools.getTransformed3DExpandedScaleHists(weightHist,
HistTools.makeUnrolledHist,
[self.unrolledBinsX, self.unrolledBinsY], name,
entries=scaleVars['entries'],
pairs=scaleVars['groups'],
)
scaleHists.extend(expandedScaleHists)
return scaleHists
def GetMousePosition(self,position,physical=False):
positionFraction = float(position[0]-self.GetPosition()[0])/float(self.GetSize()[0])
#positionPhysical = (self.xmax-self.xmin)*(positionFraction-self.fractionBinStart)/(self.fractionBinEnd-self.fractionBinStart) + self.xmin
positionPhysical = -1.
if self.isLogX:
positionPhysical = HistTools.convertLogFractionToPhysical(positionFraction,self.fractionBinStart,self.fractionBinEnd,self.xmin,self.xmax)
else:
positionPhysical = HistTools.convertLinearFractionToPhysical(positionFraction,self.fractionBinStart,self.fractionBinEnd,self.xmin,self.xmax)
if self.doCutAtBinEdge:
positionFractionTmp = positionFraction
positionFraction = self.getLowBinEdge(positionFractionTmp,True)
positionPhysical = self.getLowBinEdge(positionFractionTmp)
if physical: return positionPhysical
return positionFraction
def set_DrawDatum(xarray,
key,
nbins=100,
weights=[],
x_range=[-1, -1],
xbins=[],
color="black",
linestyle="-"):
"""
Make DrawDatum object for 1D histogram
Parameters
----------
xarray : numpy array
array of values that you want to make a histogram
key : string
name or description of the histogram, used in the label in legend
nbins : int
number of bins for the histogram
weights : numpy array
weight parameter for histogram
x_range : list
range of x axis
xbins : numpy array
if you want to give your own xbins, set it.
color : string
name of color for drawing
linestyle : string
line style marker
Returns
-------
datum : DrawDatum object
"""
datum = DrawDatum()
datum.title = key
datum.val, bins, datum.w2s = HT.make_1D_hist(xarray,
nbins=nbins,
weights=weights,
x_range=x_range,
xbins=xbins)
datum.xbinedges = bins
datum.xbins = 0.5 * (bins[1:] + bins[:-1])
datum.color = color
datum.linestyle = linestyle
if len(weights) == 0:
datum.xmean = float(sum(xarray)) / len(xarray)
else:
datum.xmean = sum(xarray * weights) / sum(weights)
return datum
def loadHistsForProcess(self,
processName,
scaleNorm=1,
expandedTheory=True):
plotsToRead = self.listOfHistsByProcess(processName,
nameReplace=("unrolled", "2D"))
group = HistTools.makeCompositeHists(
self.inputFile,
processName, {
proc: self.crossSectionMap[proc]
for proc in self.processes[processName]
},
self.lumi,
plotsToRead,
rebin=self.rebin,
overflow=False)
if self.isUnrolledFit:
for hist in group:
if not "TH2" in hist.ClassName():
continue
hist = HistTools.makeUnrolledHist(hist, self.unrolledBinsX,
self.unrolledBinsY)
histName = hist.GetName()
group.Add(hist)
fitVariable = self.getFitVariable(processName)
#TODO:Make optional
processedHists = []
for chan in self.channels:
histName = "_".join([fitVariable, chan])
hist = group.FindObject(histName)
if not hist:
logging.warning("Failed to produce hist %s for process %s" %
(histName, processName))
continue
if self.removeZeros and "data" not in processName.lower():
HistTools.removeZeros(hist)
if self.addOverflow:
HistTools.addOverflow(hist)
processedHists.append(histName)
self.yields[chan].update({
processName:
round(hist.Integral(), 3) if hist.Integral() > 0 else 0.0001
})
if chan == self.channels[0]:
self.yields["all"][processName] = self.yields[chan][
processName]
else:
self.yields["all"][processName] += self.yields[chan][
processName]
if processName in self.perbinVariations:
for varName, var, corr in self.perbinVariations[processName]:
map(
lambda x: group.Add(x),
self.perBinVariationHists(hist, varName, var, corr,
chan == self.channels[0]))
scaleHists = []
if processName in self.theoryVariations:
theoryVars = self.theoryVariations[processName]
try:
scaleHists.extend(
self.scaleHistsForProcess(group, processName, chan,
expandedTheory))
if 'theoryBasedVars' in theoryVars['scale']:
for theoryBasedVar in theoryVars['scale'][
'theoryBasedVars']:
scaleHists.extend(
self.scaleHistsForProcess(
group, processName, chan, expandedTheory,
theoryBasedVar))
except ValueError as e:
logging.warning(e)
continue
pdfHists = []
pdfVars = filter(lambda x: 'pdf' in x, theoryVars.keys())
weightHist = group.FindObject(
self.weightHistName(chan, processName))
for var in pdfVars:
pdfVar = theoryVars[var]
pdfType = "MC"
if "hessian" in pdfVar['combine']:
pdfType = "Hessian" if "assym" not in pdfVar[
'combine'] else "AssymHessian"
pdfFunction = "get%sPDFVarHists" % pdfType
args = [
weightHist, pdfVar['entries'], processName, self.rebin,
pdfVar['central'], pdfVar['name']
]
if self.isUnrolledFit:
pdfFunction = pdfFunction.replace(
"get", "getTransformed3D")
args = args[0:1] + [
HistTools.makeUnrolledHist,
[self.unrolledBinsX, self.unrolledBinsY]
] + args[1:]
updatePdfs = getattr(HistTools, pdfFunction)(*args)
pdfHists += updatePdfs
if expandedTheory and pdfVar['name'] == 'NNPDF31':
args.pop(len(args) - 1)
pdfFunction = HistTools.getAllSymHessianHists if not self.isUnrolledFit else HistTools.getTransformed3DAllSymHessianHists
allPdfHists = pdfFunction(*args)
pdfHists.extend(allPdfHists)
if not self.isUnrolledFit:
cenHist, _ = HistTools.getLHEWeightHists(
weightHist, pdfVar['entries'][:1], processName,
"pdf", self.rebin)
else:
cenHist = HistTools.getAllTransformed3DHists(
weightHist, HistTools.makeUnrolledHist,
[self.unrolledBinsX, self.unrolledBinsY],
processName, pdfVar['entries'][:1])
if len(cenHist) and cenHist[0]:
pdfHists.append(cenHist[0].Clone())
group.extend(scaleHists + pdfHists)
if chan == self.channels[0]:
theoryVarLabels = []
for h in group:
fitvar = self.getFitVariable(processName)
theoryVarLabels.extend(
re.findall("_".join([fitvar, "(.*pdf.*)", chan]),
h.GetName()))
theoryVarLabels.extend(
re.findall(
"_".join([fitvar, "(.*QCDscale.*)", chan]),
h.GetName()))
self.variations[processName].extend(set(theoryVarLabels))
if self.addOverflow:
map(HistTools.addOverflow,
filter(lambda x: (x.GetName() not in processedHists), group))
if self.removeZeros and "data" not in group.GetName().lower():
map(HistTools.removeZeros,
filter(lambda x: (x.GetName() not in processedHists), group))
#TODO: You may want to combine channels before removing zeros
if self.channelsToCombine.keys():
self.combineChannels(group, processName)
self.histData[processName] = group
def Configure(self):
nbins = 0
xmin = -1
xmax = -1
isLogX = False
if self.name=="nbjets":
nbins = 4
xmin = -.5
xmax = 3.5
isLogX = False
elif self.name=="njets":
nbins = 10
xmin = -.5
xmax = 9.5
isLogX = False
elif self.name=="pt":
nbins = 30
xmin = 20
xmax = 1000
isLogX = True
elif self.name=="phi":
nbins = 16
xmin = -math.pi
xmax = math.pi
isLogX = False
elif self.name=="eta":
nbins = 15
xmin = -2.5
xmax = 2.5
isLogX = False
elif self.name=="charge":
nbins = 7
xmin = -3.5
xmax = 3.5
isLogX = False
elif self.name=="masselel":
nbins = 20
xmin = 1.1876
xmax = 201.1876
isLogX = False
elif self.name=="masselmu":
nbins = 20
xmin = 1.1876
xmax = 201.1876
isLogX = False
elif self.name=="massmumu":
nbins = 20
xmin = 1.1876
xmax = 201.1876
isLogX = False
elif self.name=="masssfos":
nbins = 20
xmin = 1.1876
xmax = 201.1876
isLogX = False
elif self.name=="met":
nbins = 30
xmin = 20
xmax = 1000
isLogX = True
elif self.name=="metphi":
nbins = 16
xmin = -math.pi
xmax = math.pi
isLogX = False
elif self.name=="metsumet":
nbins = 30
xmin = 100.
xmax = 10000.
isLogX = True
elif self.name=="nsfos":
nbins = 3
xmin = -.5
xmax = 2.5
isLogX = False
elif self.name=="nmuons":
nbins = 4
xmin = -.5
xmax = 3.5
isLogX = False
elif self.name=="mt":
nbins = 30
xmin = 20
xmax = 1000
isLogX = True
elif self.name=="deltaphi":
nbins = 16
xmin = -math.pi
xmax = math.pi
isLogX = False
else:
print "Error! Couldn't configure feature with name",self.name
exit(2)
self.xmin = xmin
self.xmax = xmax
self.nbins = nbins
self.isLogX = isLogX
if self.isLogX:
self.binsPhysical = HistTools.getLogBins(nbins,xmin,xmax)
self.binsFraction = HistTools.getLogBins(nbins,self.fractionBinStart,self.fractionBinEnd)
else:
self.binsPhysical = HistTools.getLinearBins(nbins,xmin,xmax)
self.binsFraction = HistTools.getLinearBins(nbins,self.fractionBinStart,self.fractionBinEnd)
def loadHistsForProcess(self, processName, scaleNorm=1, expandedTheory=True):
plotsToRead = self.listOfHistsByProcess(processName)
group = HistTools.makeCompositeHists(self.inputFile, processName,
{proc : self.crossSectionMap[proc] for proc in self.processes[processName]},
self.lumi, plotsToRead, rebin=self.rebin, overflow=False)
fitVariable = self.getFitVariable(processName)
#TODO:Make optional
processedHists = []
for chan in self.channels:
histName = "_".join([fitVariable, chan]) if chan != "all" else fitVariable
hist = group.FindObject(histName)
if not hist:
raise RuntimeError("Failed to produce hist %s for process %s" % (histName, processName))
#TODO: Make optional
if "data" not in processName.lower():
HistTools.removeZeros(hist)
HistTools.addOverflow(hist)
processedHists.append(histName)
self.yields[chan].update({processName : round(hist.Integral(), 3) if hist.Integral() > 0 else 0.0001})
if chan == self.channels[0]:
self.yields["all"][processName] = self.yields[chan][processName]
else:
self.yields["all"][processName] += self.yields[chan][processName]
if processName in self.theoryVariations:
weightHist = group.FindObject(self.weightHistName(chan, processName))
if not weightHist:
logging.warning("Failed to find %s. Skipping" % self.weightHistName(chan, processName))
continue
theoryVars = self.theoryVariations[processName]
scaleHists = HistTools.getScaleHists(weightHist, processName, self.rebin,
entries=theoryVars['scale']['entries'],
central=(theoryVars['scale']['central'] if 'scale' in theoryVars else -1))
if expandedTheory:
expandedScaleHists = HistTools.getExpandedScaleHists(weightHist, processName, self.rebin,
entries=theoryVars['scale']['entries'],
central=(theoryVars['scale']['central'] if 'scale' in theoryVars else -1),
)
scaleHists.extend(expandedScaleHists)
pdfHists = []
pdfVars = filter(lambda x: 'pdf' in x, theoryVars.keys())
for var in pdfVars:
pdfVar = theoryVars[var]
pdfType = "SymmMC"
if "hessian" in pdfVar['combine']:
pdfType = "Hessian" if "assym" not in pdfVar['combine'] else "AssymHessian"
pdfFunction = "get%sPDFVariationHists" % pdfType
pdfHists += getattr(HistTools, pdfFunction)(weightHist, pdfVar['entries'], processName,
self.rebin, central=pdfVar['central'],
pdfName=pdfVar['name'])
if expandedTheory and "hessian" in pdfVar['combine']:
allPdfHists = HistTools.getAllSymmetricHessianVariationHists(weightHist, pdfVar['entries'], processName,
self.rebin, central=pdfVar['central'])
pdfHists.extend(allPdfHists)
cenHist, _ = HistTools.getLHEWeightHists(weightHist, pdfVar['entries'][:1], processName, "pdf", self.rebin)
if len(cenHist) and cenHist[0]:
pdfHists.append(cenHist[0].Clone())
group.extend(scaleHists+pdfHists)
if chan == self.channels[0]:
theoryVarLabels = []
for h in group:
theoryVarLabels.extend(re.findall("_".join([self.fitVariable, "(.*)", chan]), h.GetName()))
self.variations[processName].extend(set(theoryVarLabels))
#TODO: Make optional
map(HistTools.addOverflow, filter(lambda x: (x.GetName() not in processedHists), group))
if "data" not in group.GetName().lower():
map(HistTools.removeZeros, filter(lambda x: (x.GetName() not in processedHists), group))
#TODO: You may want to combine channels before removing zeros
self.combineChannels(group, processName)
self.histData[processName] = group
def set_DrawDatum2D(xarray,
yarray,
key,
nxbins,
nybins,
weights=[],
x_range=[-1, -1],
y_range=[-1, -1]):
"""
Make DrawDatum object for 2D histogram
Parameters
----------
xarray : numpy array
array of values for xaxis that you want to make a histogram
yarray : numpy array
array of values for yaxis that you want to make a histogram
key : string
name or description of the histogram, used in the label in legend
nxbins : int
number of bins for xaxis for the histogram
nybins : int
number of bins for yaxis for the histogram
weights : numpy array
weight parameter for histogram
x_range : list
range of x axis
y_range : list
range of y axis
Returns
-------
datum : DrawDatum object
"""
datum = DrawDatum()
datum.title = key
datum.xmeshgrid, datum.ymeshgrid, datum.val, datum.w2s, xbins, ybins = HT.make_2D_hist(
xarray,
yarray,
nxbins,
nybins,
weights=weights,
x_range=x_range,
y_range=y_range)
datum.xbinedges = datum.xmeshgrid[:, 0]
datum.xbins = 0.5 * (datum.xbinedges[1:] + datum.xbinedges[:-1])
datum.ybinedges = datum.ymeshgrid[0]
datum.ybins = 0.5 * (datum.ybinedges[1:] + datum.ybinedges[:-1])
if len(weights) == 0:
datum.xmean = float(sum(xarray)) / len(xarray)
else:
datum.xmean = sum(xarray * weights) / sum(weights)
return datum
