#bfr6.Draw()
"""
sct = RooSimWSTool(w)
model_sim = sct.build("model_sim","model",SplitParam("m","c"))
"""
mc = RooStats.ModelConfig("mc", w)
mc.SetPdf(simPdf)
mc.SetParametersOfInterest("mu")
mc.SetObservables("bdt")
mc.SetNuisanceParameters("nbkg")
mc.SetNuisanceParameters("nbkg2")
c.cd(7)
bfr7 = w.var("bdt").frame()
cdata = ROOT.RooStats.AsymptoticCalculator.MakeAsimovData(
combData, mc, w.argSet("bdt,sample"), ROOT.RooArgSet())
dh = ROOT.RooDataHist("", "", w.argSet("bdt,sample"), cdata)
err_correction = [
dh.set(dh.get(i), dh.weight(dh.get(i)),
ROOT.TMath.Sqrt(dh.weight(dh.get(i))))
for i in range(0, dh.numEntries())
]
dh.plotOn(bfr7)
bfr7.SetTitle("Asimov Dataset")
bfr7.Draw()
# h = ROOT.TH1F("hobs", "hobs", 100, -1, 1); data.fillHistogram(h, ROOT.RooArgList(w.var("bdt")))
# h.Draw()
# dh = ROOT.RooDataHist("obs", "obs", ROOT.RooArgList(w.var("bdt")), h)
getattr(w, 'import')(dh)
class Wjj2DFitter:
def __init__ (self, pars):
self.pars = pars
self.ws = RooWorkspace('wjj2dfitter')
self.utils = Wjj2DFitterUtils(self.pars)
self.useImportPars = False
self.rangeString = None
obs = []
for v in self.pars.var:
try:
vName = self.pars.varNames[v]
except AttributeError:
vName = v
obs.append(vName)
var1 = self.ws.factory('%s[%f,%f]' % (vName,
self.pars.varRanges[v][1],
self.pars.varRanges[v][2])
)
var1.setUnit('GeV')
try:
var1.SetTitle(self.pars.varTitles[v])
except AttributeError:
var1.SetTitle('m_{jj}')
var1.setPlotLabel(var1.GetTitle())
if len(self.pars.varRanges[v][3]) > 1:
vbinning = RooBinning(len(self.pars.varRanges[v][3]) - 1,
array('d', self.pars.varRanges[v][3]),
'%sBinning' % vName)
var1.setBinning(vbinning)
else:
var1.setBins(self.pars.varRanges[v][0])
var1.Print()
if v in self.pars.exclude:
var1.setRange('signalRegion', self.pars.exclude[v][0],
self.pars.exclude[v][1])
var1.setRange('lowSideband', var1.getMin(),
self.pars.exclude[v][0])
var1.setRange('highSideband', self.pars.exclude[v][1],
var1.getMax())
self.rangeString = 'lowSideband,highSideband'
if hasattr(self.pars, 'plotRanges'):
var1.setRange('plotRange', self.pars.plotRanges[v][1],
self.pars.plotRanges[v][2])
var1.setBins(self.pars.plotRanges[v][0], 'plotBins')
else:
var1.setRange('plotRange', var1.getMin(), var1.getMax())
var1.setBins(var1.getBins(), 'plotBins')
self.ws.defineSet('obsSet', ','.join(obs))
def loadDataFromWorkspace(self, other, cut = None):
#pull unbinned data from other workspace
unbinnedData = other.data('data_unbinned')
if not unbinnedData:
unbinnedData = other.data('data_obs')
if cut:
unbinnedData = unbinnedData.reduce(cut)
unbinnedData.Print()
if self.pars.binData:
#bin and import data
unbinnedData.SetName('data_unbinned')
getattr(self.ws, 'import')(unbinnedData)
data = RooDataHist('data_obs', 'data_obs', other.set('obsSet'),
unbinnedData)
getattr(self.ws, 'import')(data)
else:
#just import data
unbinnedData.SetName('data_obs')
getattr(self.ws, 'import')(unbinnedData)
def loadHistogramsFromWorkspace(self, other):
#pull RooHist pdfs from other workspace
pdfs = other.allPdfs()
pdfIter = pdfs.createIterator()
pdf = pdfIter.Next()
while pdf:
if pdf.IsA().InheritsFrom('RooHistPdf'):
print 'importing',pdf.GetName(),'from old workspace'
getattr(self.ws, 'import')(pdf)
pdf = pdfIter.Next()
def loadWorkspaceFromFile(self, filename, wsname = 'w',
getFloatPars = True):
print 'loading data workspace %s from file %s' % (wsname, filename)
fin = TFile.Open(filename)
if not fin:
print 'failed to open the file',filename
import os
print 'cwd:',os.getcwd()
print 'access of',filename,os.access(filename, os.R_OK)
print 'list of root files in cwd'
for f in os.listdir(os.getcwd()):
if f[-5:] == '.root':
print f,len(f),len(filename)
fin = TFile.Open(os.getcwd() + '/' + filename)
assert(fin)
other = fin.Get(wsname)
#pull unbinned data from other workspace
self.loadDataFromWorkspace(other)
#pull in histogram pdfs to save time
self.loadHistogramsFromWorkspace(other)
if getFloatPars and other.loadSnapshot('fitPars'):
self.useImportPars = True
self.ws.saveSnapshot('importParams', other.set('floatingParams'),
True)
# self.ws.Print()
# put together a fitting model and return the pdf
def makeFitter(self, useAlternateModels = False):
if self.ws.pdf('total'):
return self.ws.pdf('total')
compPdfs = []
for component in self.pars.backgrounds:
# print 'getting compModels'
compModels = getattr(self.pars, '%sModels' % component)
if hasattr(self.pars, '%sConvModels' % component):
convModels = getattr(self.pars, '%sConvModels' % component)
else:
convModels = None
if useAlternateModels:
print 'loading Alternate Models'
compModels = getattr(self.pars, '%sModelsAlt' % component)
convModels = getattr(self.pars, '%sConvModelsAlt' % component)
# print 'compModels = %s' % compModels
compFiles = getattr(self.pars, '%sFiles' % component)
compPdf = self.makeComponentPdf(component, compFiles, compModels,
useAlternateModels, convModels)
norm = self.ws.factory('prod::f_%s_norm' % component + \
'(n_%s[0.,1e6],' % component + \
'%s_nrm[1.,-0.5,5.])' % component)
self.ws.var('n_%s' % component).setConstant(True)
if hasattr(self, '%sExpected' % component):
self.ws.var('n_%s' % component).setVal(
getattr(self, '%sExpected' % component))
compPdfs.append(
self.ws.factory('RooExtendPdf::%s_extended(%s,%s)' % \
(compPdf.GetName(),
compPdf.GetName(),
norm.GetName())
)
)
self.ws.factory('r_signal[0., -200., 200.]')
self.ws.var('r_signal').setConstant(False)
try:
obs = [ self.pars.varNames[x] for x in self.pars.var ]
except AttributeError:
obs = self.pars.var
for component in self.pars.signals:
compFile = getattr(self.pars, '%sFiles' % component)
compModels = getattr(self.pars, '%sModels' % component)
if hasattr(self.pars, '%sConvModels' % component):
convModels = getattr(self.pars, '%sConvModels' % component)
else:
convModels = None
compPdf = self.makeComponentPdf(component, compFiles, compModels,
useAlternateModels, convModels)
norm = self.ws.factory(
"prod::f_%s_norm(n_%s[0., 1e6],r_signal)" % \
(component, component)
)
self.ws.var('n_%s' % component).setConstant(True)
if hasattr(self, '%sExpected' % component):
self.ws.var('n_%s' % component).setVal(
getattr(self, '%sExpected' % component))
pdf = self.ws.factory('RooExtendPdf::%s_extended(%s,%s)' % \
(compPdf.GetName(),
compPdf.GetName(),
norm.GetName())
)
if (hasattr(self.pars, '%sInterference' % component)) and \
getattr(self.pars, '%sInterference' % component):
getattr(self.ws, 'import') \
(pdf, RooFit.RenameAllNodes('interf_%sUp' % component),
RooFit.RenameAllVariablesExcept('interf_%sUp' % component,
','.join(obs)),
RooFit.Silence()
)
getattr(self.ws, 'import') \
(pdf, RooFit.RenameAllNodes('interf_%sDown' % component),
RooFit.RenameAllVariablesExcept('interf_%sDown'%component,
','.join(obs)),
RooFit.Silence()
)
if self.pars.includeSignal:
compPdfs.append(pdf)
#print compPdfs
prodList = [ '%s' % (pdf.GetName()) \
for (idx, pdf) in enumerate(compPdfs) ]
comps = RooArgList(self.ws.argSet(','.join(prodList)))
getattr(self.ws, 'import')(RooAddPdf('total', 'total', comps))
return self.ws.pdf('total')
# define the constraints on the yields, etc that will be part of the fit.
def makeConstraints(self):
if self.ws.set('constraintSet'):
return self.ws.set('constraintSet')
constraints = []
constrainedParameters = []
for constraint in self.pars.yieldConstraints:
theYield = self.ws.var('%s_nrm' % constraint)
if not theYield.isConstant():
self.ws.factory('RooGaussian::%s_const(%s, 1.0, %f)' % \
(constraint, theYield.GetName(),
self.pars.yieldConstraints[constraint])
)
constraints.append('%s_const' % constraint)
constrainedParameters.append(theYield.GetName())
if hasattr(self.pars, 'constrainShapes'):
for component in self.pars.constrainShapes:
pc = self.ws.pdf(component).getParameters(self.ws.set('obsSet'))
parIter = pc.createIterator()
par = parIter.Next()
while par:
if not par.isConstant():
theConst = self.ws.factory('RooGaussian::%s_const' % \
(par.GetName()) + \
'(%s, %f, %f)' % \
(par.GetName(),
par.getVal(),
par.getError())
)
constraints.append(theConst.GetName())
constrainedParameters.append(par.GetName())
par = parIter.Next()
pc.IsA().Destructor(pc)
self.ws.defineSet('constraintSet', ','.join(constraints))
self.ws.defineSet('constrainedSet', ','.join(constrainedParameters))
return self.ws.set('constraintSet')
# make the constrained fitter
def makeConstrainedFitter(self):
if self.ws.pdf('totalFit_const'):
return self.ws.pdf('totalFit_const')
constraintSet = self.makeConstraints()
fitter = self.makeFitter()
print '\nfit constraints'
constIter = constraintSet.createIterator()
constraint = constIter.Next()
constraints = []
while constraint:
constraint.Print()
constraints.append(constraint.GetName())
constraint = constIter.Next()
if constraintSet.getSize() > 0:
constraints.append(fitter.GetName())
fitter = self.ws.factory('PROD::totalFit_const(%s)' % \
(','.join(constraints))
)
return fitter
# fit the data using the pdf
def fit(self, keepParameterValues = False, overrideRangeCmd = False):
print 'construct fit pdf ...'
fitter = self.makeFitter()
print 'load data ...'
data = self.loadData()
self.resetYields()
constraintSet = self.makeConstraints()
if not keepParameterValues:
self.readParametersFromFile()
self.resetYields()
# print constraints, self.pars.yieldConstraints
constraintCmd = RooCmdArg.none()
if constraintSet.getSize() > 0:
fitter = self.makeConstrainedFitter()
constraintCmd = RooFit.Constrained()
# constraintCmd = RooFit.ExternalConstraints(self.ws.set('constraintSet'))
if self.useImportPars:
self.ws.loadSnapshot('importParams')
self.ws.Print()
# for constraint in pars.constraints:
# self.ws.pdf(constraint).Print()
# print
rangeCmd = RooCmdArg.none()
if self.rangeString and self.pars.doExclude and not overrideRangeCmd:
rangeCmd = RooFit.Range(self.rangeString)
# print 'scanning parameter values...'
# fitter.fitTo(data, RooFit.Minos(False),
# RooFit.PrintEvalErrors(-1),
# RooFit.Warnings(False),
# RooFit.Minimizer("Minuit2", "scan"),
# RooFit.PrintLevel(0),
# constraintCmd,
# rangeCmd)
print 'fitting ...'
fr = fitter.fitTo(data, RooFit.Save(True),
# RooFit.Extended(True),
RooFit.Minos(False),
RooFit.PrintEvalErrors(-1),
RooFit.Warnings(False),
RooFit.Minimizer("Minuit2", "minimize"),
constraintCmd,
rangeCmd
)
fr.Print('v')
return fr
# determine the fitting model for each component and return them
def makeComponentPdf(self, component, files, models, useAlternateModels,
convModels):
print 'making ComponentPdf %s' % component
# print 'models = %s' % models
# print 'files = %s' % files
if convModels and not (convModels[0] == -1):
thePdf = self.makeConvolvedPdf(component, files, models, useAlternateModels, convModels)
elif (models[0] == -1):
thePdf = self.makeComponentHistPdf(component, files)
elif (models[0] == -2):
thePdf = self.makeMorphingPdf(component, useAlternateModels, convModels)
elif (models[0] == -3):
pass
else:
thePdf = self.makeComponentAnalyticPdf(component, models, useAlternateModels)
return thePdf
#create a simple 2D histogram pdf
def makeComponentHistPdf(self, component, files):
if self.ws.pdf(component):
return self.ws.pdf(component)
compHist = self.utils.newEmptyHist('hist%s' % component)
sumYields = 0.
sumxsec = 0.
sumExpected = 0.
for (idx,fset) in enumerate(files):
if hasattr(self.pars, '%scuts' % component):
cutOverride = getattr(self.pars, '%scuts' % component)
else:
cutOverride = None
filename = fset[0]
tmpHist = self.utils.File2Hist(filename,
'hist%s_%i' % (component, idx),
False,cutOverride,False,True,0)
sumYields += tmpHist.Integral()
sumxsec += fset[2]
compHist.Add(tmpHist, self.pars.integratedLumi*fset[2]/fset[1])
sumExpected += tmpHist.Integral()*fset[2]* \
self.pars.integratedLumi/fset[1]
print filename,'acc x eff: %.3g' % (tmpHist.Integral()/fset[1])
print filename,'N_expected: %.1f' % \
(tmpHist.Integral()*fset[2]*self.pars.integratedLumi/fset[1])
#tmpHist.Print()
#compHist.Print()
print '%s acc x eff: %.3g' % \
(component, sumExpected/sumxsec/self.pars.integratedLumi)
print 'Number of expected %s events: %.1f' % (component, sumExpected)
setattr(self, '%sExpected' % component, sumExpected)
return self.utils.Hist2Pdf(compHist, component,
self.ws, self.pars.order)
#create a pdf which is a convolution of any two pdf
def makeConvolvedPdf(self, component, files, models, useAlternateModels, convModels):
if self.ws.pdf(component):
return self.ws.pdf(component)
#If a morphing model is selected, then convolve each individual component first and then morph
if (models[0] == -2):
return self.makeMorphingPdf(component, useAlternateModels, convModels)
basePdf = self.makeComponentPdf('%s_base' % component, files, models, useAlternateModels, [-1])
convComponent = 'Global' ##Overwrite to use the same convolution model for all Pdfs
convModel = getattr(self.pars, '%sConvModels' % convComponent)
if useAlternateModels:
convModel = getattr(self.pars, '%sConvModelsAlt' % convComponent)
convPdf = self.makeComponentPdf('%s_conv' % convComponent, files, convModel, useAlternateModels, [-1])
var = self.pars.var[0]
try:
vName = self.pars.varNames[var]
except AttributeError:
vName = var
self.ws.factory('RooFFTConvPdf::%s(%s,%s,%s)' % \
(component, vName, basePdf.GetName(),
convPdf.GetName()))
return self.ws.pdf(component)
# create a pdf using the "template morphing" technique
def makeMorphingPdf(self, component, useAlternateModels, convModels):
if self.ws.pdf(component):
return self.ws.pdf(component)
filesNom = getattr(self.pars, '%s_NomFiles' % component)
modelsNom = getattr(self.pars, '%s_NomModels' % component)
filesMU = getattr(self.pars, '%s_MUFiles' % component)
modelsMU = getattr(self.pars, '%s_MUModels' % component)
filesMD = getattr(self.pars, '%s_MDFiles' % component)
modelsMD = getattr(self.pars, '%s_MDModels' % component)
filesSU = getattr(self.pars, '%s_SUFiles' % component)
modelsSU = getattr(self.pars, '%s_SUModels' % component)
filesSD = getattr(self.pars, '%s_SDFiles' % component)
modelsSD = getattr(self.pars, '%s_SDModels' % component)
if useAlternateModels:
modelsNom = getattr(self.pars, '%s_NomModelsAlt' % component)
modelsMU = getattr(self.pars, '%s_MUModelsAlt' % component)
modelsMD = getattr(self.pars, '%s_MDModelsAlt' % component)
modelsSU = getattr(self.pars, '%s_SUModelsAlt' % component)
modelsSD = getattr(self.pars, '%s_SDModelsAlt' % component)
# Adds five (sub)components for the component with suffixes Nom, MU, MD, SU, SD
NomPdf = self.makeComponentPdf('%s_Nom' % component, filesNom, modelsNom, False, convModels)
if hasattr(self, '%s_NomExpected' % component):
setattr(self, '%sExpected' % component,
getattr(self, '%s_NomExpected' % component))
MUPdf = self.makeComponentPdf('%s_MU' % component, filesMU, modelsMU, False, convModels)
MDPdf = self.makeComponentPdf('%s_MD' % component, filesMD, modelsMD, False, convModels)
SUPdf = self.makeComponentPdf('%s_SU' % component, filesSU, modelsSU, False, convModels)
SDPdf = self.makeComponentPdf('%s_SD' % component, filesSD, modelsSD, False, convModels)
fMU_comp = self.ws.factory("fMU_%s[0., -1., 1.]" % component)
fSU_comp = self.ws.factory("fSU_%s[0., -1., 1.]" % component)
fMU = RooFormulaVar("f_fMU_%s" % component, "1.0*@0*(@0 >= 0.)",
RooArgList( fMU_comp ) )
fMD = RooFormulaVar("f_fMD_%s" % component, "-1.0*@0*(@0 < 0.)",
RooArgList( fMU_comp ) )
fSU = RooFormulaVar("f_fSU_%s" % component, "@0*(@0 >= 0.)",
RooArgList( fSU_comp ) )
fSD = RooFormulaVar("f_fSD_%s" % component, "@0*(-1)*(@0 < 0.)",
RooArgList( fSU_comp ) )
fNom = RooFormulaVar("f_fNom_%s" % component, "(1.-abs(@0)-abs(@1))",
RooArgList(fMU_comp,fSU_comp) )
morphPdf = RooAddPdf(component,component,
RooArgList(MUPdf,MDPdf,SUPdf,SDPdf,NomPdf),
RooArgList(fMU, fMD, fSU, fSD, fNom))
morphPdf.SetName(component)
getattr(self.ws, 'import')(morphPdf)
return self.ws.pdf(component)
# create a pdf using an analytic function.
def makeComponentAnalyticPdf(self, component, models, useAlternateModels):
if self.ws.pdf(component):
return self.ws.pdf(component)
pdfList = []
systMult = None
if ( hasattr(self.pars, '%sInterference' % component) and \
getattr(self.pars, '%sInterference' % component) and \
hasattr(self.pars, "%sdoSystMult" % component) and \
getattr(self.pars, "%sdoSystMult" % component) ):
systMult = getattr(self.pars, "%sSystMult" % component)
for (idx,model) in enumerate(models):
var = self.pars.var[idx]
try:
vName = self.pars.varNames[var]
except AttributeError:
vName = var
auxModel = None
if useAlternateModels:
if hasattr(self.pars, '%sAuxModelsAlt' % component):
auxModel = getattr(self.pars, '%sAuxModelsAlt' % component)[idx]
else:
if hasattr(self.pars, '%sAuxModels' % component):
auxModel = getattr(self.pars, '%sAuxModels' % component)[idx]
pdfList.append(self.utils.analyticPdf(self.ws, vName, model,
'%s_%s'%(component,vName),
'%s_%s'%(component,vName),
auxModel, systMult
)
)
pdfListNames = [ pdf.GetName() for pdf in pdfList ]
if len(pdfList) > 1:
self.ws.factory('PROD::%s(%s)' % \
(component, ','.join(pdfListNames)))
else:
pdfList[0].SetName(component)
return self.ws.pdf(component)
def loadData(self, weight = False):
if self.ws.data('data_obs'):
return self.ws.data('data_obs')
unbinnedName = 'data_obs'
if self.pars.binData:
unbinnedName = 'data_unbinned'
data = self.utils.File2Dataset(self.pars.DataFile, unbinnedName,
self.ws, weighted = weight)
if self.pars.binData:
data = RooDataHist('data_obs', 'data_obs', self.ws.set('obsSet'),
data)
getattr(self.ws, 'import')(data)
data = self.ws.data('data_obs')
return data
def stackedPlot(self, var, logy = False, pdfName = None, Silent = False):
if not pdfName:
pdfName = 'total'
xvar = self.ws.var(var)
nbins = xvar.getBins()
# if hasattr(self.pars, 'plotRanges') and not xvar.hasRange('plotRange'):
# xvar.setRange('plotRange', self.pars.plotRanges[var][1],
# self.pars.plotRanges[var][2])
# xvar.setBins(self.pars.plotRanges[var][0], 'plotBins')
# elif not xvar.hasRange('plotRange'):
# xvar.setRange('plotRange', xvar.getMin(), xvar.getMax())
# xvar.setBins(nbins, 'plotBins')
sframe = xvar.frame(RooFit.Range('plotRange'),
RooFit.Bins(xvar.getBins('plotBins')))
sframe.SetName("%s_stacked" % var)
pdf = self.ws.pdf(pdfName)
if isinstance(pdf, RooAddPdf):
compList = RooArgList(pdf.pdfList())
else:
compList = None
data = self.ws.data('data_obs')
nexp = pdf.expectedEvents(self.ws.set('obsSet'))
if not Silent:
print pdf.GetName(),'expected: %.0f' % (nexp)
print 'data events: %.0f' % (data.sumEntries())
if nexp < 1:
nexp = data.sumEntries()
theComponents = []
if self.pars.includeSignal:
theComponents += self.pars.signals
theComponents += self.pars.backgrounds
data.plotOn(sframe, RooFit.Invisible(),
RooFit.Binning('plotBins'))
# dataHist = RooAbsData.createHistogram(data,'dataHist_%s' % var, xvar,
# RooFit.Binning('%sBinning' % var))
# #dataHist.Scale(1., 'width')
# invData = RooHist(dataHist, 1., 1, RooAbsData.SumW2, 1.0, False)
# #invData.Print('v')
# sframe.addPlotable(invData, 'pe', True, True)
for (idx,component) in enumerate(theComponents):
if not Silent:
print 'plotting',component,'...',
if hasattr(self.pars, '%sPlotting' % (component)):
plotCharacteristics = getattr(self.pars, '%sPlotting' % \
(component))
else:
plotCharacteristics = {'color' : colorwheel[idx%6],
'title' : component }
compCmd = RooCmdArg.none()
if compList:
compSet = RooArgSet(compList)
if compSet.getSize() > 0:
compCmd = RooFit.Components(compSet)
removals = compList.selectByName('%s*' % component)
compList.remove(removals)
if not Silent:
print 'events', self.ws.function('f_%s_norm' % component).getVal()
sys.stdout.flush()
if abs(self.ws.function('f_%s_norm' % component).getVal()) >= 1.:
pdf.plotOn(sframe, #RooFit.ProjWData(data),
RooFit.DrawOption('LF'), RooFit.FillStyle(1001),
RooFit.FillColor(plotCharacteristics['color']),
RooFit.LineColor(plotCharacteristics['color']),
RooFit.VLines(),
RooFit.Range('plotRange'),
RooFit.NormRange('plotRange'),
RooFit.Normalization(nexp, RooAbsReal.NumEvent),
compCmd
)
tmpCurve = sframe.getCurve()
tmpCurve.SetName(component)
tmpCurve.SetTitle(plotCharacteristics['title'])
if 'visible' in plotCharacteristics:
sframe.setInvisible(component,
plotCharacteristics['visible'])
data.plotOn(sframe, RooFit.Name('theData'),
RooFit.Binning('plotBins'))
sframe.getHist('theData').SetTitle('data')
# theData = RooHist(dataHist, 1., 1, RooAbsData.SumW2, 1.0, True)
# theData.SetName('theData')
# theData.SetTitle('data')
# sframe.addPlotable(theData, 'pe')
if (logy):
sframe.SetMinimum(0.01)
sframe.SetMaximum(1.0e6)
else:
sframe.SetMaximum(sframe.GetMaximum()*1.35)
pass
excluded = (var in self.pars.exclude)
bname = var
if not excluded:
for v in self.pars.exclude:
if hasattr(self.pars, 'varNames') and \
(self.pars.varNames[v] == var):
excluded = True
bname = v
if excluded:
blinder = TBox(self.pars.exclude[bname][0], sframe.GetMinimum(),
self.pars.exclude[bname][1], sframe.GetMaximum())
# blinder.SetName('blinder')
# blinder.SetTitle('signal region')
blinder.SetFillColor(kBlack)
if self.pars.blind:
blinder.SetFillStyle(1001)
else:
blinder.SetFillStyle(0)
blinder.SetLineStyle(2)
sframe.addObject(blinder)
elif self.pars.blind:
if not Silent:
print "blind but can't find exclusion region for", var
print 'excluded',excluded,self.pars.exclude
print 'hiding data points'
sframe.setInvisible('theData', True)
else:
sframe.setInvisible('theData', False)
#sframe.GetYaxis().SetTitle('Events / GeV')
# dataHist.IsA().Destructor(dataHist)
if not Silent:
print
xvar.setBins(nbins)
return sframe
def readParametersFromFile(self, fname=None):
if (not fname):
fname = self.pars.initialParametersFile
if isinstance(fname, str):
flist = [ fname ]
else:
flist = fname
for tmpName in flist:
if len(tmpName) > 0:
print 'loading parameters from file',tmpName
self.ws.allVars().readFromFile(tmpName)
def expectedFromPars(self):
components = self.pars.signals + self.pars.backgrounds
for component in components:
theYield = self.ws.var('n_%s' % component)
setattr(self, '%sExpected' % component, theYield.getVal())
def initFromExplicitVals(self,opts):
#,init_diboson= -1.0,init_WpJ=-1.0,init_top=-1.0,init_ZpJ=-1.0,init_QCD=-1.0
components = ['diboson', 'top', 'WpJ', 'ZpJ', 'QCD', 'WHbb']
for component in components:
#double init
init = getattr(opts, 'ext%s' % component)
#init = -2.0
#setattr(self,init, 'init_%s' % component)
#init = init_%s % component
#print "init=", init
#init = self.ws.var('init_%s' % component)
#init.setVal(100.0)
#init.setVal('init_%s' % component)
#init = theYield.getVal()
if (init>0.):
print 'setting initial value for ',component,' to ',init
setattr(self, '%sInitial' % component, init)
def resetYields(self):
if self.ws.data('data_obs'):
Ndata = self.ws.data('data_obs').sumEntries()
else:
Ndata = 10000.
print 'resetting yields...'
components = self.pars.signals + self.pars.backgrounds
for component in components:
theYield = self.ws.var('n_%s' % component)
theNorm = self.ws.var('%s_nrm' % component)
if hasattr(self, '%sInitial' % component):
print 'explicitly setting initial value for ',component
theYield.setVal(getattr(self, '%sInitial' % component))
theNorm.setVal(1.0)
theNorm.setConstant()
else:
fracofdata = -1.
if hasattr(self.pars, '%sFracOfData' % component):
fracofdata = getattr(self.pars, '%sFracOfData' % component)
if (fracofdata >= 0.):
print 'explicitly setting ', component,' yield to be', fracofdata,' of data'
theYield.setVal(fracofdata*Ndata)
elif hasattr(self, '%sExpected' % component):
theYield.setVal(getattr(self, '%sExpected' % component))
else:
print 'no expected value for',component
theYield.setVal(Ndata/len(components))
if theNorm and not theNorm.isConstant():
theNorm.setVal(1.0)
if component in self.pars.yieldConstraints:
theYield.setError(theYield.getVal() * \
self.pars.yieldConstraints[component])
if theNorm:
theNorm.setError(self.pars.yieldConstraints[component])
else:
theYield.setError(sqrt(theYield.getVal()))
theYield.Print()
def generateToyMCSet(self,var,inputPdf,outFileName,NEvts):
fMC = TFile(outFileName, "RECREATE");
# thevar = self.ws.var(var);
print 'thevar='
print var
# print thevar
print '...'
# varList = RooArgList()
# varList.add(self.ws.var(var))
toymc = inputPdf.generate(RooArgSet(self.ws.var(var)),NEvts);
tMC = toymc.tree();
fMC.cd();
tMC.Write();
fMC.Close();
def legend4Plot(plot, left = False):
if left:
theLeg = TLegend(0.2, 0.62, 0.55, 0.92, "", "NDC")
else:
theLeg = TLegend(0.60, 0.62, 0.92, 0.92, "", "NDC")
theLeg.SetName('theLegend')
theLeg.SetBorderSize(0)
theLeg.SetLineColor(0)
theLeg.SetFillColor(0)
theLeg.SetFillStyle(0)
theLeg.SetLineWidth(0)
theLeg.SetLineStyle(0)
theLeg.SetTextFont(42)
theLeg.SetTextSize(.045)
entryCnt = 0
for obj in range(0, int(plot.numItems())):
objName = plot.nameOf(obj)
if (not plot.getInvisible(objName)):
theObj = plot.getObject(obj)
objTitle = theObj.GetTitle()
if len(objTitle) < 1:
objTitle = objName
dopts = plot.getDrawOptions(objName).Data()
# print 'obj:',theObj,'title:',objTitle,'opts:',dopts,'type:',type(dopts)
if theObj.IsA().InheritsFrom('TNamed'):
theLeg.AddEntry(theObj, objTitle, dopts)
entryCnt += 1
theLeg.SetY1NDC(0.9 - 0.05*entryCnt - 0.005)
theLeg.SetY1(theLeg.GetY1NDC())
return theLeg
legend4Plot = staticmethod(legend4Plot)
norm = combinedWS.function('f_%s_norm' % comp)
combinedWS.factory('PROD::%s(%s_%s, %s)' % (comp, comp, other_var,
newPdf.GetName()))
compPdfs.append(combinedWS.factory('RooExtendPdf::%s_extended(%s, %s)' %\
(comp, comp, norm.GetName())))
combinedWS.defineSet('obsSet', '%s,%s' % (other_var, var_name))
if opts.sigInject:
combinedWS.var('r_signal').setVal(opts.sigInject)
combinedWS.var('r_signal').setError(0.1)
combinedWS.var('r_signal').setRange(-3., 9.)
combinedWS.var('r_signal').setConstant(False)
compNames = [ c.GetName() for c in compPdfs ]
compList = RooArgList(combinedWS.argSet(','.join(compNames)))
getattr(combinedWS, 'import')(RooAddPdf('total', 'total', compList))
combinedPdf = combinedWS.pdf('total')
genPars = combinedPdf.getParameters(combinedWS.set('obsSet'))
combinedWS.defineSet('params', genPars)
if opts.toy:
#generate toy dataset
print 'Generated parameters'
combinedWS.set('params').Print('v')
combinedWS.saveSnapshot("genPars", combinedWS.set('params'))
data = combinedPdf.generate(combinedWS.set('obsSet'),
RooFit.Name('data_obs'),
RooFit.Extended())
data.Print('v')
w.obj("bkg_only").plotOn(f, RooFit.LineColor(4))
f.SetTitle("Generated Dataset")
f.Draw()
mc = RooStats.ModelConfig("mc", w)
mc.SetPdf(w.pdf("model"))
mc.SetParametersOfInterest("mu")
mc.SetObservables("bdt")
mc.SetNuisanceParameters("nbkg")
#getattr(w, 'import')(mc)
c.cd(3)
bfr2 = w.var("bdt").frame()
data = ROOT.RooStats.AsymptoticCalculator.MakeAsimovData(
data, mc, ROOT.RooArgSet(w.var("bdt")), ROOT.RooArgSet())
dh = ROOT.RooDataHist("", "", w.argSet("bdt"), data)
err_correction = [
dh.set(dh.get(i), dh.weight(dh.get(i)),
ROOT.TMath.Sqrt(dh.weight(dh.get(i))))
for i in range(0, dh.numEntries())
]
dh.plotOn(bfr2)
bfr2.SetTitle("Asimov Dataset")
bfr2.Draw()
# exit(0)
# h = ROOT.TH1F("hobs", "hobs", 100, -1, 1); data.fillHistogram(h, ROOT.RooArgList(w.var("bdt")))
# h.Draw()
# dh = ROOT.RooDataHist("obs", "obs", ROOT.RooArgList(w.var("bdt")), h)
getattr(w, 'import')(dh)
class Wjj2DFitter:
def __init__ (self, pars):
self.pars = pars
self.ws = RooWorkspace('wjj2dfitter')
self.utils = Wjj2DFitterUtils(self.pars)
self.useImportPars = False
self.rangeString = None
obs = []
for v in self.pars.var:
try:
vName = self.pars.varNames[v]
except AttributeError:
vName = v
obs.append(vName)
var1 = self.ws.factory('%s[%f,%f]' % (vName,
self.pars.varRanges[v][1],
self.pars.varRanges[v][2])
)
var1.setUnit('GeV')
try:
var1.SetTitle(self.pars.varTitles[v])
except AttributeError:
var1.SetTitle('m_{jj}')
var1.setPlotLabel(var1.GetTitle())
if len(self.pars.varRanges[v][3]) > 1:
vbinning = RooBinning(len(self.pars.varRanges[v][3]) - 1,
array('d', self.pars.varRanges[v][3]),
'%sBinning' % vName)
var1.setBinning(vbinning)
else:
var1.setBins(self.pars.varRanges[v][0])
var1.Print()
if v in self.pars.exclude:
var1.setRange('signalRegion', self.pars.exclude[v][0],
self.pars.exclude[v][1])
var1.setRange('lowSideband', var1.getMin(),
self.pars.exclude[v][0])
var1.setRange('highSideband', self.pars.exclude[v][1],
var1.getMax())
self.rangeString = 'lowSideband,highSideband'
self.ws.defineSet('obsSet', ','.join(obs))
def loadDataFromWorkspace(self, other, cut = None):
#pull unbinned data from other workspace
unbinnedData = other.data('data_unbinned')
if not unbinnedData:
unbinnedData = other.data('data_obs')
if cut:
unbinnedData = unbinnedData.reduce(cut)
unbinnedData.Print()
if self.pars.binData:
#bin and import data
unbinnedData.SetName('data_unbinned')
getattr(self.ws, 'import')(unbinnedData)
data = RooDataHist('data_obs', 'data_obs', other.set('obsSet'),
unbinnedData)
getattr(self.ws, 'import')(data)
else:
#just import data
unbinnedData.SetName('data_obs')
getattr(self.ws, 'import')(unbinnedData)
def loadHistogramsFromWorkspace(self, other):
#pull RooHist pdfs from other workspace
pdfs = other.allPdfs()
pdfIter = pdfs.createIterator()
pdf = pdfIter.Next()
while pdf:
if pdf.IsA().InheritsFrom('RooHistPdf'):
print 'importing',pdf.GetName(),'from old workspace'
getattr(self.ws, 'import')(pdf)
pdf = pdfIter.Next()
def loadWorkspaceFromFile(self, filename, wsname = 'w',
getFloatPars = True):
print 'loading data workspace %s from file %s' % (wsname, filename)
fin = TFile.Open(filename)
if not fin:
print 'failed to open the file',filename
import os
print 'cwd:',os.getcwd()
print 'access of',filename,os.access(filename, os.R_OK)
print 'list of root files in cwd'
for f in os.listdir(os.getcwd()):
if f[-5:] == '.root':
print f,len(f),len(filename)
fin = TFile.Open(os.getcwd() + '/' + filename)
assert(fin)
other = fin.Get(wsname)
#pull unbinned data from other workspace
self.loadDataFromWorkspace(other)
#pull in histogram pdfs to save time
self.loadHistogramsFromWorkspace(other)
if getFloatPars and other.loadSnapshot('fitPars'):
self.useImportPars = True
self.ws.saveSnapshot('importParams', other.set('floatingParams'),
True)
# self.ws.Print()
# put together a fitting model and return the pdf
def makeFitter(self, useAlternateModels = False):
if self.ws.pdf('total'):
return self.ws.pdf('total')
compPdfs = []
for component in self.pars.backgrounds:
# print 'getting compModels'
compModels = getattr(self.pars, '%sModels' % component)
if hasattr(self.pars, '%sConvModels' % component):
convModels = getattr(self.pars, '%sConvModels' % component)
else:
convModels = None
if useAlternateModels:
print 'loading Alternate Models'
compModels = getattr(self.pars, '%sModelsAlt' % component)
convModels = getattr(self.pars, '%sConvModelsAlt' % component)
# print 'compModels = %s' % compModels
compFiles = getattr(self.pars, '%sFiles' % component)
compPdf = self.makeComponentPdf(component, compFiles, compModels,
useAlternateModels, convModels)
norm = self.ws.factory('prod::f_%s_norm' % component + \
'(n_%s[0.,1e6],' % component + \
'%s_nrm[1.,-0.5,5.])' % component)
self.ws.var('n_%s' % component).setConstant(True)
if hasattr(self, '%sExpected' % component):
self.ws.var('n_%s' % component).setVal(
getattr(self, '%sExpected' % component))
compPdfs.append(
self.ws.factory('RooExtendPdf::%s_extended(%s,%s)' % \
(compPdf.GetName(),
compPdf.GetName(),
norm.GetName())
)
)
self.ws.factory('r_signal[0., -200., 200.]')
self.ws.var('r_signal').setConstant(False)
try:
obs = [ self.pars.varNames[x] for x in self.pars.var ]
except AttributeError:
obs = self.pars.var
for component in self.pars.signals:
compFile = getattr(self.pars, '%sFiles' % component)
compModels = getattr(self.pars, '%sModels' % component)
if hasattr(self.pars, '%sConvModels' % component):
convModels = getattr(self.pars, '%sConvModels' % component)
else:
convModels = None
compPdf = self.makeComponentPdf(component, compFiles, compModels,
useAlternateModels, convModels)
norm = self.ws.factory(
"prod::f_%s_norm(n_%s[0., 1e6],r_signal)" % \
(component, component)
)
self.ws.var('n_%s' % component).setConstant(True)
if hasattr(self, '%sExpected' % component):
self.ws.var('n_%s' % component).setVal(
getattr(self, '%sExpected' % component))
pdf = self.ws.factory('RooExtendPdf::%s_extended(%s,%s)' % \
(compPdf.GetName(),
compPdf.GetName(),
norm.GetName())
)
if (hasattr(self.pars, '%sInterference' % component)) and \
getattr(self.pars, '%sInterference' % component):
getattr(self.ws, 'import') \
(pdf, RooFit.RenameAllNodes('interf_%sUp' % component),
RooFit.RenameAllVariablesExcept('interf_%sUp' % component,
','.join(obs)),
RooFit.Silence()
)
getattr(self.ws, 'import') \
(pdf, RooFit.RenameAllNodes('interf_%sDown' % component),
RooFit.RenameAllVariablesExcept('interf_%sDown'%component,
','.join(obs)),
RooFit.Silence()
)
if self.pars.includeSignal:
compPdfs.append(pdf)
#print compPdfs
prodList = [ '%s' % (pdf.GetName()) \
for (idx, pdf) in enumerate(compPdfs) ]
comps = RooArgList(self.ws.argSet(','.join(prodList)))
getattr(self.ws, 'import')(RooAddPdf('total', 'total', comps))
return self.ws.pdf('total')
# define the constraints on the yields, etc that will be part of the fit.
def makeConstraints(self):
if self.ws.set('constraintSet'):
return self.ws.set('constraintSet')
constraints = []
constrainedParameters = []
for constraint in self.pars.yieldConstraints:
theYield = self.ws.var('%s_nrm' % constraint)
if not theYield.isConstant():
self.ws.factory('RooGaussian::%s_const(%s, 1.0, %f)' % \
(constraint, theYield.GetName(),
self.pars.yieldConstraints[constraint])
)
constraints.append('%s_const' % constraint)
constrainedParameters.append(theYield.GetName())
if hasattr(self.pars, 'constrainShapes'):
for component in self.pars.constrainShapes:
pc = self.ws.pdf(component).getParameters(self.ws.set('obsSet'))
parIter = pc.createIterator()
par = parIter.Next()
while par:
if not par.isConstant():
theConst = self.ws.factory('RooGaussian::%s_const' % \
(par.GetName()) + \
'(%s, %f, %f)' % \
(par.GetName(),
par.getVal(),
par.getError())
)
constraints.append(theConst.GetName())
constrainedParameters.append(par.GetName())
par = parIter.Next()
pc.IsA().Destructor(pc)
self.ws.defineSet('constraintSet', ','.join(constraints))
self.ws.defineSet('constrainedSet', ','.join(constrainedParameters))
return self.ws.set('constraintSet')
# fit the data using the pdf
def fit(self, keepParameterValues = False):
print 'construct fit pdf ...'
fitter = self.makeFitter()
print 'load data ...'
data = self.loadData()
self.resetYields()
constraintSet = self.makeConstraints()
if not keepParameterValues:
self.readParametersFromFile()
self.resetYields()
# print constraints, self.pars.yieldConstraints
print '\nfit constraints'
constIter = constraintSet.createIterator()
constraint = constIter.Next()
constraints = []
while constraint:
constraint.Print()
constraints.append(constraint.GetName())
constraint = constIter.Next()
constraintCmd = RooCmdArg.none()
if constraintSet.getSize() > 0:
constraints.append(fitter.GetName())
fitter = self.ws.pdf('totalFit_const')
if not fitter:
fitter = self.ws.factory('PROD::totalFit_const(%s)' % \
(','.join(constraints))
)
constraintCmd = RooFit.Constrained()
# constraintCmd = RooFit.ExternalConstraints(self.ws.set('constraintSet'))
if self.useImportPars:
self.ws.loadSnapshot('importParams')
self.ws.Print()
# for constraint in pars.constraints:
# self.ws.pdf(constraint).Print()
# print
rangeCmd = RooCmdArg.none()
if self.rangeString and self.pars.doExclude:
rangeCmd = RooFit.Range(self.rangeString)
print 'fitting ...'
fr = fitter.fitTo(data, RooFit.Save(True),
RooFit.Extended(True),
RooFit.Minos(False),
RooFit.PrintEvalErrors(-1),
RooFit.Warnings(False),
constraintCmd,
rangeCmd)
fr.Print()
return fr
# determine the fitting model for each component and return them
def makeComponentPdf(self, component, files, models, useAlternateModels,
convModels):
print 'making ComponentPdf %s' % component
# print 'models = %s' % models
# print 'files = %s' % files
if convModels and not (convModels[0] == -1):
thePdf = self.makeConvolvedPdf(component, files, models, useAlternateModels, convModels)
elif (models[0] == -1):
thePdf = self.makeComponentHistPdf(component, files)
elif (models[0] == -2):
thePdf = self.makeMorphingPdf(component, useAlternateModels, convModels)
elif (models[0] == -3):
pass
else:
thePdf = self.makeComponentAnalyticPdf(component, models, useAlternateModels)
return thePdf
#create a simple 2D histogram pdf
def makeComponentHistPdf(self, component, files):
if self.ws.pdf(component):
return self.ws.pdf(component)
compHist = self.utils.newEmptyHist('hist%s' % component)
sumYields = 0.
sumxsec = 0.
sumExpected = 0.
for (idx,fset) in enumerate(files):
if hasattr(self.pars, '%scuts' % component):
cutOverride = getattr(self.pars, '%scuts' % component)
else:
cutOverride = None
filename = fset[0]
tmpHist = self.utils.File2Hist(filename,
'hist%s_%i' % (component, idx),
False,cutOverride,False,True,0)
sumYields += tmpHist.Integral()
sumxsec += fset[2]
compHist.Add(tmpHist, self.pars.integratedLumi*fset[2]/fset[1])
sumExpected += tmpHist.Integral()*fset[2]* \
self.pars.integratedLumi/fset[1]
print filename,'acc x eff: %.3g' % (tmpHist.Integral()/fset[1])
print filename,'N_expected: %.1f' % \
(tmpHist.Integral()*fset[2]*self.pars.integratedLumi/fset[1])
#tmpHist.Print()
#compHist.Print()
print '%s acc x eff: %.3g' % \
(component, sumExpected/sumxsec/self.pars.integratedLumi)
print 'Number of expected %s events: %.1f' % (component, sumExpected)
setattr(self, '%sExpected' % component, sumExpected)
return self.utils.Hist2Pdf(compHist, component,
self.ws, self.pars.order)
#create a pdf which is a convolution of any two pdf
def makeConvolvedPdf(self, component, files, models, useAlternateModels, convModels):
if self.ws.pdf(component):
return self.ws.pdf(component)
#If a morphing model is selected, then convolve each individual component first and then morph
if (models[0] == -2):
return self.makeMorphingPdf(component, useAlternateModels, convModels)
basePdf = self.makeComponentPdf('%s_base' % component, files, models, useAlternateModels, [-1])
convComponent = 'Global' ##Overwrite to use the same convolution model for all Pdfs
convModel = getattr(self.pars, '%sConvModels' % convComponent)
if useAlternateModels:
convModel = getattr(self.pars, '%sConvModelsAlt' % convComponent)
convPdf = self.makeComponentPdf('%s_conv' % convComponent, files, convModel, useAlternateModels, [-1])
var = self.pars.var[0]
try:
vName = self.pars.varNames[var]
except AttributeError:
vName = var
self.ws.factory('RooFFTConvPdf::%s(%s,%s,%s)' % \
(component, vName, basePdf.GetName(),
convPdf.GetName()))
return self.ws.pdf(component)
# create a pdf using the "template morphing" technique
def makeMorphingPdf(self, component, useAlternateModels, convModels):
if self.ws.pdf(component):
return self.ws.pdf(component)
filesNom = getattr(self.pars, '%s_NomFiles' % component)
modelsNom = getattr(self.pars, '%s_NomModels' % component)
filesMU = getattr(self.pars, '%s_MUFiles' % component)
modelsMU = getattr(self.pars, '%s_MUModels' % component)
filesMD = getattr(self.pars, '%s_MDFiles' % component)
modelsMD = getattr(self.pars, '%s_MDModels' % component)
filesSU = getattr(self.pars, '%s_SUFiles' % component)
modelsSU = getattr(self.pars, '%s_SUModels' % component)
filesSD = getattr(self.pars, '%s_SDFiles' % component)
modelsSD = getattr(self.pars, '%s_SDModels' % component)
if useAlternateModels:
modelsNom = getattr(self.pars, '%s_NomModelsAlt' % component)
modelsMU = getattr(self.pars, '%s_MUModelsAlt' % component)
modelsMD = getattr(self.pars, '%s_MDModelsAlt' % component)
modelsSU = getattr(self.pars, '%s_SUModelsAlt' % component)
modelsSD = getattr(self.pars, '%s_SDModelsAlt' % component)
# Adds five (sub)components for the component with suffixes Nom, MU, MD, SU, SD
NomPdf = self.makeComponentPdf('%s_Nom' % component, filesNom, modelsNom, False, convModels)
if hasattr(self, '%s_NomExpected' % component):
setattr(self, '%sExpected' % component,
getattr(self, '%s_NomExpected' % component))
MUPdf = self.makeComponentPdf('%s_MU' % component, filesMU, modelsMU, False, convModels)
MDPdf = self.makeComponentPdf('%s_MD' % component, filesMD, modelsMD, False, convModels)
SUPdf = self.makeComponentPdf('%s_SU' % component, filesSU, modelsSU, False, convModels)
SDPdf = self.makeComponentPdf('%s_SD' % component, filesSD, modelsSD, False, convModels)
fMU_comp = self.ws.factory("fMU_%s[0., -1., 1.]" % component)
fSU_comp = self.ws.factory("fSU_%s[0., -1., 1.]" % component)
fMU = RooFormulaVar("f_fMU_%s" % component, "1.0*@0*(@0 >= 0.)",
RooArgList( fMU_comp ) )
fMD = RooFormulaVar("f_fMD_%s" % component, "-1.0*@0*(@0 < 0.)",
RooArgList( fMU_comp ) )
fSU = RooFormulaVar("f_fSU_%s" % component, "@0*(@0 >= 0.)",
RooArgList( fSU_comp ) )
fSD = RooFormulaVar("f_fSD_%s" % component, "@0*(-1)*(@0 < 0.)",
RooArgList( fSU_comp ) )
fNom = RooFormulaVar("f_fNom_%s" % component, "(1.-abs(@0)-abs(@1))",
RooArgList(fMU_comp,fSU_comp) )
morphPdf = RooAddPdf(component,component,
RooArgList(MUPdf,MDPdf,SUPdf,SDPdf,NomPdf),
RooArgList(fMU, fMD, fSU, fSD, fNom))
morphPdf.SetName(component)
getattr(self.ws, 'import')(morphPdf)
return self.ws.pdf(component)
# create a pdf using an analytic function.
def makeComponentAnalyticPdf(self, component, models, useAlternateModels):
if self.ws.pdf(component):
return self.ws.pdf(component)
pdfList = []
for (idx,model) in enumerate(models):
var = self.pars.var[idx]
try:
vName = self.pars.varNames[var]
except AttributeError:
vName = var
auxModel = None
if useAlternateModels:
if hasattr(self.pars, '%sAuxModelsAlt' % component):
auxModel = getattr(self.pars, '%sAuxModelsAlt' % component)[idx]
else:
if hasattr(self.pars, '%sAuxModels' % component):
auxModel = getattr(self.pars, '%sAuxModels' % component)[idx]
pdfList.append(self.utils.analyticPdf(self.ws, vName, model,
'%s_%s'%(component,vName),
'%s_%s'%(component,vName),
auxModel
)
)
pdfListNames = [ pdf.GetName() for pdf in pdfList ]
if len(pdfList) > 1:
self.ws.factory('PROD::%s(%s)' % \
(component, ','.join(pdfListNames)))
else:
pdfList[0].SetName(component)
return self.ws.pdf(component)
def loadData(self, weight = False):
if self.ws.data('data_obs'):
return self.ws.data('data_obs')
unbinnedName = 'data_obs'
if self.pars.binData:
unbinnedName = 'data_unbinned'
data = self.utils.File2Dataset(self.pars.DataFile, unbinnedName,
self.ws, weighted = weight)
if self.pars.binData:
data = RooDataHist('data_obs', 'data_obs', self.ws.set('obsSet'),
data)
getattr(self.ws, 'import')(data)
data = self.ws.data('data_obs')
return data
def stackedPlot(self, var, logy = False, pdfName = None, Silent = False):
if not pdfName:
pdfName = 'total'
xvar = self.ws.var(var)
nbins = xvar.getBins()
if hasattr(self.pars, 'plotRanges'):
xvar.setRange('plotRange', self.pars.plotRanges[var][1],
self.pars.plotRanges[var][2])
xvar.setBins(self.pars.plotRanges[var][0], 'plotBins')
else:
xvar.setRange('plotRange', xvar.getMin(), xvar.getMax())
xvar.setBins(nbins, 'plotBins')
sframe = xvar.frame()
sframe.SetName("%s_stacked" % var)
pdf = self.ws.pdf(pdfName)
if isinstance(pdf, RooAddPdf):
compList = RooArgList(pdf.pdfList())
else:
compList = None
data = self.ws.data('data_obs')
nexp = pdf.expectedEvents(self.ws.set('obsSet'))
if not Silent:
print pdf.GetName(),'expected: %.0f' % (nexp)
print 'data events: %.0f' % (data.sumEntries())
if nexp < 1:
nexp = data.sumEntries()
theComponents = []
if self.pars.includeSignal:
theComponents += self.pars.signals
theComponents += self.pars.backgrounds
data.plotOn(sframe, RooFit.Invisible(),
RooFit.Binning('plotBins'))
# dataHist = RooAbsData.createHistogram(data,'dataHist_%s' % var, xvar,
# RooFit.Binning('%sBinning' % var))
# #dataHist.Scale(1., 'width')
# invData = RooHist(dataHist, 1., 1, RooAbsData.SumW2, 1.0, False)
# #invData.Print('v')
# sframe.addPlotable(invData, 'pe', True, True)
for (idx,component) in enumerate(theComponents):
if not Silent:
print 'plotting',component,'...',
if hasattr(self.pars, '%sPlotting' % (component)):
plotCharacteristics = getattr(self.pars, '%sPlotting' % \
(component))
else:
plotCharacteristics = {'color' : colorwheel[idx%6],
'title' : component }
compCmd = RooCmdArg.none()
if compList:
compSet = RooArgSet(compList)
if compSet.getSize() > 0:
compCmd = RooFit.Components(compSet)
removals = compList.selectByName('%s*' % component)
compList.remove(removals)
if not Silent:
print 'events', self.ws.function('f_%s_norm' % component).getVal()
sys.stdout.flush()
if abs(self.ws.function('f_%s_norm' % component).getVal()) >= 1.:
pdf.plotOn(sframe, #RooFit.ProjWData(data),
RooFit.DrawOption('LF'), RooFit.FillStyle(1001),
RooFit.FillColor(plotCharacteristics['color']),
RooFit.LineColor(plotCharacteristics['color']),
RooFit.VLines(),
RooFit.Range('plotRange'),
RooFit.NormRange('plotRange'),
RooFit.Normalization(nexp, RooAbsReal.NumEvent),
compCmd
)
tmpCurve = sframe.getCurve()
tmpCurve.SetName(component)
tmpCurve.SetTitle(plotCharacteristics['title'])
if 'visible' in plotCharacteristics:
sframe.setInvisible(component,
plotCharacteristics['visible'])
data.plotOn(sframe, RooFit.Name('theData'),
RooFit.Binning('plotBins'))
sframe.getHist('theData').SetTitle('data')
# theData = RooHist(dataHist, 1., 1, RooAbsData.SumW2, 1.0, True)
# theData.SetName('theData')
# theData.SetTitle('data')
# sframe.addPlotable(theData, 'pe')
if (logy):
sframe.SetMinimum(0.01)
sframe.SetMaximum(1.0e6)
else:
sframe.SetMaximum(sframe.GetMaximum()*1.35)
pass
excluded = (var in self.pars.exclude)
bname = var
if not excluded:
for v in self.pars.exclude:
if hasattr(self.pars, 'varNames') and \
(self.pars.varNames[v] == var):
excluded = True
bname = v
if excluded:
blinder = TBox(self.pars.exclude[bname][0], sframe.GetMinimum(),
self.pars.exclude[bname][1], sframe.GetMaximum())
# blinder.SetName('blinder')
# blinder.SetTitle('signal region')
blinder.SetFillColor(kBlack)
if self.pars.blind:
blinder.SetFillStyle(1001)
else:
blinder.SetFillStyle(0)
blinder.SetLineStyle(2)
sframe.addObject(blinder)
elif self.pars.blind:
if not Silent:
print "blind but can't find exclusion region for", var
print 'excluded',excluded,self.pars.exclude
print 'hiding data points'
sframe.setInvisible('theData', True)
#sframe.GetYaxis().SetTitle('Events / GeV')
# dataHist.IsA().Destructor(dataHist)
if not Silent:
print
xvar.setBins(nbins)
return sframe
def readParametersFromFile(self, fname=None):
if (not fname):
fname = self.pars.initialParametersFile
if isinstance(fname, str):
flist = [ fname ]
else:
flist = fname
for tmpName in flist:
if len(tmpName) > 0:
print 'loading parameters from file',tmpName
self.ws.allVars().readFromFile(tmpName)
def expectedFromPars(self):
components = self.pars.signals + self.pars.backgrounds
for component in components:
theYield = self.ws.var('n_%s' % component)
setattr(self, '%sExpected' % component, theYield.getVal())
def initFromExplicitVals(self,opts):
#,init_diboson= -1.0,init_WpJ=-1.0,init_top=-1.0,init_ZpJ=-1.0,init_QCD=-1.0
components = ['diboson', 'top', 'WpJ', 'ZpJ', 'QCD', 'WHbb']
for component in components:
#double init
init = getattr(opts, 'ext%s' % component)
#init = -2.0
#setattr(self,init, 'init_%s' % component)
#init = init_%s % component
#print "init=", init
#init = self.ws.var('init_%s' % component)
#init.setVal(100.0)
#init.setVal('init_%s' % component)
#init = theYield.getVal()
if (init>0.):
print 'setting initial value for ',component,' to ',init
setattr(self, '%sInitial' % component, init)
def resetYields(self):
if self.ws.data('data_obs'):
Ndata = self.ws.data('data_obs').sumEntries()
else:
Ndata = 10000.
print 'resetting yields...'
components = self.pars.signals + self.pars.backgrounds
for component in components:
theYield = self.ws.var('n_%s' % component)
theNorm = self.ws.var('%s_nrm' % component)
if hasattr(self, '%sInitial' % component):
print 'explicitly setting initial value for ',component
theYield.setVal(getattr(self, '%sInitial' % component))
theNorm.setVal(1.0)
theNorm.setConstant()
else:
fracofdata = -1.
if hasattr(self.pars, '%sFracOfData' % component):
fracofdata = getattr(self.pars, '%sFracOfData' % component)
if (fracofdata >= 0.):
print 'explicitly setting ', component,' yield to be', fracofdata,' of data'
theYield.setVal(fracofdata*Ndata)
elif hasattr(self, '%sExpected' % component):
theYield.setVal(getattr(self, '%sExpected' % component))
else:
print 'no expected value for',component
theYield.setVal(Ndata/len(components))
if theNorm and not theNorm.isConstant():
theNorm.setVal(1.0)
if component in self.pars.yieldConstraints:
theYield.setError(theYield.getVal() * \
self.pars.yieldConstraints[component])
if theNorm:
theNorm.setError(self.pars.yieldConstraints[component])
else:
theYield.setError(sqrt(theYield.getVal()))
theYield.Print()
def generateToyMCSet(self,var,inputPdf,outFileName,NEvts):
fMC = TFile(outFileName, "RECREATE");
# thevar = self.ws.var(var);
print 'thevar='
print var
# print thevar
print '...'
# varList = RooArgList()
# varList.add(self.ws.var(var))
toymc = inputPdf.generate(RooArgSet(self.ws.var(var)),NEvts);
tMC = toymc.tree();
fMC.cd();
tMC.Write();
fMC.Close();
def legend4Plot(plot, left = False):
if left:
theLeg = TLegend(0.2, 0.62, 0.55, 0.92, "", "NDC")
else:
theLeg = TLegend(0.60, 0.62, 0.92, 0.92, "", "NDC")
theLeg.SetName('theLegend')
theLeg.SetBorderSize(0)
theLeg.SetLineColor(0)
theLeg.SetFillColor(0)
theLeg.SetFillStyle(0)
theLeg.SetLineWidth(0)
theLeg.SetLineStyle(0)
theLeg.SetTextFont(42)
theLeg.SetTextSize(.045)
entryCnt = 0
for obj in range(0, int(plot.numItems())):
objName = plot.nameOf(obj)
if (not plot.getInvisible(objName)):
theObj = plot.getObject(obj)
objTitle = theObj.GetTitle()
if len(objTitle) < 1:
objTitle = objName
dopts = plot.getDrawOptions(objName).Data()
# print 'obj:',theObj,'title:',objTitle,'opts:',dopts,'type:',type(dopts)
if theObj.IsA().InheritsFrom('TNamed'):
theLeg.AddEntry(theObj, objTitle, dopts)
entryCnt += 1
theLeg.SetY1NDC(0.9 - 0.05*entryCnt - 0.005)
theLeg.SetY1(theLeg.GetY1NDC())
return theLeg
legend4Plot = staticmethod(legend4Plot)
from ROOT import RooFit, RooWorkspace, RooDataSet, kDashed, TBrowser
w = RooWorkspace("w", True)
w.factory("Gaussian::gauss(mes[5.20,5.30],mean[5.28,5.2,5.3],width[0.0027,0.001,1])")
w.factory("ArgusBG::argus(mes,5.291,argpar[-20,-100,-1])")
w.factory("SUM::sum(nsig[200,0,10000]*gauss,nbkg[800,0,10000]*argus)")
#--- Generate a toyMC sample from composite PDF ---
data = w.function('sum').generate(w.argSet('mes'), 2000)
#--- Perform extended ML fit of composite PDF to toy data ---
w.function('sum').fitTo(data)
# --- Plot toy data and composite PDF overlaid ---
mesframe = w.var('mes').frame()
data.plotOn(mesframe)
w.function('sum').plotOn(mesframe)
w.function('sum').plotOn(mesframe, RooFit.Components('argus'), RooFit.LineStyle(kDashed))
mesframe.Draw()
mesframe.Browse(TBrowser())
print 'nsig:',w.var('nsig').getValV(), '+-', w.var('nsig').getError()
print 'nbkg:', w.var('nbkg').getValV(), '+-', w.var('nbkg').getError()
print 'mes:', w.var('mes').getValV(), '+-', w.var('mes').getError()
print 'mean:', w.var('mean').getValV(), '+-', w.var('mean').getError()
print 'width:', w.var('width').getValV(), '+-', w.var('width').getError()
print 'argpar:', w.var('argpar').getValV(), '+-', w.var('argpar').getError()
from time import sleep
sleep(5)
from ROOT import RooFit, RooWorkspace, RooDataSet, kDashed, TBrowser
w = RooWorkspace("w", True)
w.factory(
"Gaussian::gauss(mes[5.20,5.30],mean[5.28,5.2,5.3],width[0.0027,0.001,1])")
w.factory("ArgusBG::argus(mes,5.291,argpar[-20,-100,-1])")
w.factory("SUM::sum(nsig[200,0,10000]*gauss,nbkg[800,0,10000]*argus)")
#--- Generate a toyMC sample from composite PDF ---
data = w.function('sum').generate(w.argSet('mes'), 2000)
#--- Perform extended ML fit of composite PDF to toy data ---
w.function('sum').fitTo(data)
# --- Plot toy data and composite PDF overlaid ---
mesframe = w.var('mes').frame()
data.plotOn(mesframe)
w.function('sum').plotOn(mesframe)
w.function('sum').plotOn(mesframe, RooFit.Components('argus'),
RooFit.LineStyle(kDashed))
mesframe.Draw()
mesframe.Browse(TBrowser())
print 'nsig:', w.var('nsig').getValV(), '+-', w.var('nsig').getError()
print 'nbkg:', w.var('nbkg').getValV(), '+-', w.var('nbkg').getError()
print 'mes:', w.var('mes').getValV(), '+-', w.var('mes').getError()
print 'mean:', w.var('mean').getValV(), '+-', w.var('mean').getError()
print 'width:', w.var('width').getValV(), '+-', w.var('width').getError()
print 'argpar:', w.var('argpar').getValV(), '+-', w.var('argpar').getError()
from time import sleep
sleep(5)
