def __init__( self, ParNames, AnaFiles = [ ], RefFile = '' ) :
assert ParNames and hasattr( ParNames, '__iter__' ) and hasattr( ParNames, '__len__' ) and len(ParNames) > 0, 'P2VV'\
'P2VV - ERROR: fitResultsAnalysis: no parameter names specified'
self._parNames = ParNames
from P2VV.Parameterizations.FullPDFs import PdfConfiguration
self._refPars = PdfConfiguration()
self._anaPars = PdfConfiguration()
self._refParVals = [ None ] * len(self._parNames)
self._anaParVals = { }
self._parHists = { }
self._nFiles = 0
self.readRefFile(RefFile)
self.readAnaFiles(AnaFiles)
massPdfPars = sWeightMassPdf.getParameters(preDataSet)
if model == 'ipatia2' :
# set values of constant Ipatia 2 parameters
from P2VV.Utilities.General import getSplitPar
for splitCat in sWeightMassPdf.indexCat() :
sWeightMassPdf.indexCat().setIndex( splitCat.getVal() )
KKState = observables['KKMassCat'].getLabel()
for it, parName in enumerate(constSplitPars) :
par = getSplitPar( parName, KKState, massPdfPars )
par.setVal( constSplitVals[KKState][it] )
if parFileIn or parFileOut :
# create PDF configuration object
from P2VV.Parameterizations.FullPDFs import PdfConfiguration as PdfConfig
pdfConfig = PdfConfig()
print 120 * '='
if parFileIn :
# read parameters from file
pdfConfig.readParametersFromFile( filePath = parFileIn )
pdfConfig.setParametersInPdf(sWeightMassPdf)
print 'P2VV - INFO: createBs2JpsiKKDataSet: observables in mass PDF:'
massPdfObs.Print('v')
print 'P2VV - INFO: createBs2JpsiKKDataSet: parameters in mass PDF:'
massPdfPars.Print('v')
# determine mass parameters with a fit
print 'P2VV - INFO: createBs2JpsiKKDataSet: fitting mass PDF'
simMassFitResult = sWeightMassPdf.fitTo( preDataSet, Save = True, SumW2Error = False, NumCPU = args.numCPU, Optimize = 1, Timer = True
, Minimizer = 'Minuit2', Strategy = 1, Offset = True, Hesse = runHesse, Minos = minosPars )
pdf = SimultaneousPdf( protoPdf.GetName() + '_simul', MasterPdf = protoPdf
, SplitCategories = splitCats, SplitParameters = splitPars )
# set values of split parameters
for par, vals in tResPars.iteritems() :
for cat, val in vals.iteritems() : ws[ par + '_' + cat ].setVal(val)
for par, vals in SWaveAmps.iteritems() :
for cat, val in vals.iteritems() : ws[ par + '_' + cat ].setVal(val)
else :
pdf = protoPdf
if parFileIn :
# read parameter values from file
from P2VV.Parameterizations.FullPDFs import PdfConfiguration
pdfConfig = PdfConfiguration()
pdfConfig.readParametersFromFile( filePath = parFileIn )
pdfConfig.setParametersInPdf(pdf)
###########################################################################################################################################
## compute angular efficiency moments ##
########################################
# print PDF, data, variables and parameters
print '\nData set:'
data.Print()
if normPdf :
print '\nPDF:'
pdf.Print()
if valPos < 0 :
print 'plotNLL: ERROR: file "%s" is not an NLL file for parameter "%s"' % ( filePath, testParName )
if parValsNLL :
minNLL = min(nllVals)
parValsNLLArr = array( 'd', parValsNLL )
nllValsArr = array( 'd', [ val - minNLL for val in nllVals ] )
# get fit values
parValsFit = [ ]
nllValsFit = [ ]
from P2VV.Parameterizations.FullPDFs import PdfConfiguration
for filePath in fitFilePaths :
pdfConfig = PdfConfiguration()
fitStatus = pdfConfig.readParametersFromFile( filePath, Verbose = False )
if fitStatus[0] != 0 :
print 'plotNLL: WARNING: fit status is "%d" for fit "%s"' % ( fitStatus[0], filePath )
assert testParName in pdfConfig.parameters(), 'plotNLL: ERROR: "%s" not in parameters of file "%s"' % ( testParName, filePath )
parVal = pdfConfig.parameters()[testParName][0]
pos = 0
for it in range( len(parValsFit) ) :
if parVal > parValsFit[it] :
break
pos += 1
parValsFit.insert( pos, parVal )
nllValsFit.insert( pos, fitStatus[1] )
print parValsFit
class fitResultsAnalysis(object) :
def __init__( self, ParNames, AnaFiles = [ ], RefFile = '' ) :
assert ParNames and hasattr( ParNames, '__iter__' ) and hasattr( ParNames, '__len__' ) and len(ParNames) > 0, 'P2VV'\
'P2VV - ERROR: fitResultsAnalysis: no parameter names specified'
self._parNames = ParNames
from P2VV.Parameterizations.FullPDFs import PdfConfiguration
self._refPars = PdfConfiguration()
self._anaPars = PdfConfiguration()
self._refParVals = [ None ] * len(self._parNames)
self._anaParVals = { }
self._parHists = { }
self._nFiles = 0
self.readRefFile(RefFile)
self.readAnaFiles(AnaFiles)
def readRefFile( self, file ) :
self._refParVals = [ None ] * len(self._parNames)
if file :
self._refPars.readParametersFromFile( filePath = file )
parVals = self._refPars.parameters()
for parIt, name in enumerate(self._parNames) :
assert name in parVals\
, 'P2VV - ERROR: fitResultsAnalysis.readRefFile(): parameter "%s" not found in file "%s"' % ( name, file )
self._refParVals[parIt] = ( parVals[name][0], parVals[name][1] )
def readAnaFiles( self, files ) :
self._anaParVals = dict( [ ( name, [ ] ) for name in self._parNames ] )
self._nFiles = 0
if files :
fileStats = { }
for file in files :
fitStatus = self._anaPars.readParametersFromFile( filePath = file, Verbose = False )
if fitStatus[0] :
print 'P2VV - WARNING: fitResultsAnalysis.readAnaFiles(): fit status %d for file "%s"' % ( fitStatus[0], file )
if not fitStatus[0] in fileStats :
fileStats[ fitStatus[0] ] = 1
else :
fileStats[ fitStatus[0] ] += 1
parVals = self._anaPars.parameters()
for name, vals in self._anaParVals.iteritems() :
assert name in parVals\
, 'P2VV - ERROR: fitResultsAnalysis.readAnaFiles(): parameter "%s" not found in file "%s"' % ( name, file )
vals.append( parVals[name][ : 2 ] )
self._nFiles = sum( stats for stats in fileStats.itervalues() )
print 'P2VV - INFO: fitResultsAnalysis.readAnaFiles(): read %d parameter files for analysis (fit status: %s)'\
% ( self._nFiles, ', '.join( '%d: %d' % ( stat, count ) for stat, count in fileStats.iteritems() ) )
def processResults( self, **kwargs ) :
if not self._anaParVals or self._nFiles == 0 :
print 'P2VV - WARNING: fitResultsAnalysis.processResults(): no parameter values available for analysis'
return
histsFile = kwargs.pop( 'HistsFile', '' ).split('.')[0]
anaParSetts = kwargs.pop( 'ParSettings', { } )
toyMode = kwargs.pop( 'ToyMode', False )
self._nParVals = [ [ 0, 0, 0 ] for name in self._parNames ]
self._maxPulls = [ 0. for name in self._parNames ]
self._parVals = [ [ ] for name in self._parNames ]
self._parSums = [ [ 0., 0. ] for name in self._parNames ]
self._errSums = [ 0. for name in self._parNames ]
self._pullVals = [ [ ] for name in self._parNames ]
self._pullSums = [ [ 0., 0., 0., 0., 0., 0. ] for name in self._parNames ]
if histsFile :
from P2VV.Load import LHCbStyle
from ROOT import TH1D, TLine, TCanvas, kFullDotLarge, kBlue
dCanv = TCanvas('dummy')
dCanv.Print( histsFile + '_pars.pdf[' )
dCanv.Print( histsFile + '_pulls.pdf[' )
def drawHist( parName, parHist, fileName, doFit = False, refVal = None ) :
print 'P2VV - INFO: fitResultsAnalysis.processResults(): drawing "%s" histogram' % parName
parSetts = anaParSetts[parName] if parName in anaParSetts else { }
self._parHists[parName] = parHist
parHist.GetXaxis().SetNdivisions( 7, 5, 0 )
parHist.SetTitleOffset( 1.15, 'x' )
parHist.SetTitleOffset( 1.05, 'y' )
parHist.SetXTitle( parSetts['name'] if 'name' in parSetts else parName )
parHist.SetYTitle('Entries / %.2g' % parHist.GetBinWidth(1) )
parHist.SetMarkerStyle(kFullDotLarge)
parHist.SetMarkerSize(0.7)
parHist.SetLineWidth(2)
if doFit :
parHist.Fit('gaus')
fitFunc = parHist.GetFunction('gaus')
fitFunc.SetNpx(1000)
fitFunc.SetLineWidth(3)
fitFunc.SetLineColor(kBlue)
histMax = -1.
for bin in range( 1, parHist.GetNbinsX() + 1 ) :
val = parHist.GetBinContent(bin) + parHist.GetBinErrorUp(bin)
if val > histMax : histMax = val
assert histMax > 0.
histMax *= 1.05
parHist.SetMinimum(0.)
parHist.SetMaximum(histMax)
canv = TCanvas(parName)
canv.SetLeftMargin(0.19)
canv.SetRightMargin(0.05)
canv.SetBottomMargin(0.21)
canv.SetTopMargin(0.05)
parHist.Draw('E1')
if refVal != None :
line = TLine()
line.DrawLine( refVal, 0., refVal, histMax )
canv.Print(fileName)
for parIt, name in enumerate(self._parNames) :
isPhase = name in anaParSetts and 'isPhase' in anaParSetts[name] and anaParSetts[name]['isPhase']
from math import floor, pi
for vals in self._anaParVals[name] :
self._nParVals[parIt][0] += 1
val = vals[0]
if self._refParVals[parIt] != None :
diff = val - self._refParVals[parIt][0]
if isPhase and abs(diff) >= pi :
nShift = floor( 0.5 * ( diff / pi + 1. ) )
val -= nShift * 2. * pi
diff -= nShift * 2. * pi
pull = diff / ( vals[1] if toyMode else self._refParVals[parIt][1] )
self._nParVals[parIt][ 2 if pull < 0. else 1 ] += 1
if abs(pull) > self._maxPulls[parIt] : self._maxPulls[parIt] = abs(pull)
self._pullVals[parIt].append(pull)
self._pullSums[parIt][ 1 if pull < 0. else 0 ] += pull
self._pullSums[parIt][ 3 if pull < 0. else 2 ] += pull**2
self._pullSums[parIt][ 5 if pull < 0. else 4 ] += pull**4
self._parVals[parIt].append(val)
self._parSums[parIt][0] += val
self._parSums[parIt][1] += val**2
self._errSums[parIt] += vals[1]
if name in self._parHists and self._parHists[name] :
self._parHists[name].Delete()
if histsFile :
parMinMax = anaParSetts[name]['parMinMax'] if name in anaParSetts and 'parMinMax' in anaParSetts[name] else None
parHistBins = anaParSetts[name]['parHistBins'] if name in anaParSetts and 'parHistBins' in anaParSetts[name]\
else int( float(self._nParVals[parIt][0]) / 250. ) if self._nParVals[parIt][0] > 1000 else 10
plotShift = anaParSetts[name]['plotShift'] if name in anaParSetts and 'plotShift' in anaParSetts[name] else 0.
parHistMin = parMinMax[0] if parMinMax else min( self._parVals[parIt] ) + plotShift
parHistMax = parMinMax[1] if parMinMax else max( self._parVals[parIt] ) + plotShift
parHistRange = parHistMax - parHistMin
if not parMinMax and parHistRange > 0. :
parHistMin = parHistMin - 0.01 * parHistRange
parHistMax = parHistMax + 0.01 * parHistRange
parHist = TH1D( name + '_par', name, parHistBins, parHistMin, parHistMax )
for val in self._parVals[parIt] : parHist.Fill( val + plotShift )
doParFit = name in anaParSetts and 'doParFit' in anaParSetts[name] and anaParSetts[name]['doParFit']
drawHist( name, parHist, histsFile + '_pars.pdf', doParFit
, self._refParVals[parIt][0] + plotShift if self._refParVals[parIt] != None else None )
pullMinMax = anaParSetts[name]['pullMinMax'] if name in anaParSetts and 'pullMinMax' in anaParSetts[name] else None
pullHistBins = anaParSetts[name]['pullHistBins'] if name in anaParSetts and 'pullHistBins' in anaParSetts[name]\
else int( float(self._nParVals[parIt][0]) / 250. ) if self._nParVals[parIt][0] > 1000 else 10
pullHistMin = pullMinMax[0] if pullMinMax else min( self._pullVals[parIt] )
pullHistMax = pullMinMax[1] if pullMinMax else max( self._pullVals[parIt] )
pullHistRange = pullHistMax - pullHistMin
if not pullMinMax and pullHistRange > 0. :
pullHistMin = pullHistMin - 0.01 * pullHistRange
pullHistMax = pullHistMax + 0.01 * pullHistRange
pullHist = TH1D( name + '_pull', name, pullHistBins, pullHistMin, pullHistMax )
for val in self._pullVals[parIt] : pullHist.Fill(val)
doPullFit = name in anaParSetts and 'doPullFit' in anaParSetts[name] and anaParSetts[name]['doPullFit']
drawHist(name, pullHist, histsFile + '_pulls.pdf', doPullFit, 0. )
if histsFile :
dCanv.Print( histsFile + '_pars.pdf]' )
dCanv.Print( histsFile + '_pulls.pdf]' )
for valCounts in self._nParVals :
assert valCounts[0] == self._nFiles and ( valCounts[1] + valCounts[2] == self._nFiles or self._refParVals[0] == None )
from math import sqrt, log10, ceil
nameLen = min( 30, max( len(name) for name in self._parNames ) )
sepStr = ' ' + '-' * ( nameLen + ( 167 if self._refParVals[0] != None else 32 ) )
print 'P2VV - INFO: fitResultsAnalysis.processResults(): parameter statistics for %d files:' % self._nFiles
print sepStr
print ( ' {0:<%ds} {1:<22s}' % nameLen ).format( 'name', 'measured' ),
if self._refParVals[0] != None :
print ' {0:<18s} {1:<20s} {2:<19s} {3:<19s} {4:<19s} {5:<6s} {6:<6s} {7:<8s}'\
.format( 'reference', 'pull mean', 'mean error', 'pull RMS', 'width error', '+RMS', '-RMS', 'max pull' )
else :
print
print sepStr
for name, nParVals, maxPull, parSums, errSums, pullSums, refVal\
in zip( self._parNames, self._nParVals, self._maxPulls, self._parSums, self._errSums, self._pullSums, self._refParVals ) :
meanVal = parSums[0] / float( nParVals[0] )
meanSqVal = parSums[1] / float( nParVals[0] )
stdDev = sqrt( meanSqVal - meanVal**2 )
precDev = max( 0, 3 - int( ceil( log10(stdDev) ) ) )
prec = max( 0, 3 - int( ceil( log10( refVal[1] ) ) ) ) if refVal != None else precDev
print ( ' {0:<%ds} {1:<+8.%df} {2:<11.%df}' % ( nameLen, prec, precDev ) ).format( name, meanVal, stdDev ),
if refVal != None :
refErr = errSums / float( nParVals[0] ) if toyMode else refVal[1]
pullMean = ( pullSums[0] + pullSums[1] ) / float( nParVals[0] )
pullVar = ( pullSums[2] + pullSums[3] ) / float( nParVals[0] )
pullVarVar = ( pullSums[4] + pullSums[5] ) / float( nParVals[0] )
meanErr = sqrt( ( pullVar - pullMean**2 ) / float( nParVals[0] ) )
width = sqrt(pullVar)
widthErr = 0.5 / sqrt(pullVar) * sqrt( ( pullVarVar - pullVar**2 ) / float( nParVals[0] ) )
widthPlus = sqrt( pullSums[2] / float( nParVals[1] ) ) if nParVals[1] > 0 else 0.
widthMin = sqrt( pullSums[3] / float( nParVals[2] ) ) if nParVals[2] > 0 else 0.
print ( ' {0:<+8.%df} {1:<7.%df} {2:<+6.3f} / {3:<+11.%df} {4:<5.3f} / {5:<11.%df} {6:<5.3f} / {7:<11.%df} {8:<5.3f} / {9:<11.%df} {10:<6.3f} {11:<6.3f} {12:<8.3f}'\
% ( prec, prec, precDev, precDev, precDev, precDev ) )\
.format( refVal[0], refErr, pullMean, pullMean * refErr, meanErr, meanErr * refErr
, width, width * refErr, widthErr, widthErr * refErr, widthPlus, widthMin, maxPull )
else :
print
print sepStr
print 13 * ' ' + 'split parameters:'
for it, pars in enumerate(splitPars) :
print 13 * ' ' + '%2d: pars: [ %s ]' % ( it, ', '.join( par.GetName() for par in pars[0] ) )
print 13 * ' ' + ' cats: [ %s ]' % ', '.join( cat.GetName() for cat in pars[1] )
from P2VV.RooFitWrappers import SimultaneousPdf
simMassPdf = SimultaneousPdf( massPdf.GetName() + '_simul'
, MasterPdf = massPdf
, SplitCategories = [ pars[1] for pars in splitPars ]
, SplitParameters = [ pars[0] for pars in splitPars ]
)
massPdfObs = simMassPdf.getObservables(dataSet)
massPdfPars = simMassPdf.getParameters(dataSet)
# create PDF configuration object
from P2VV.Parameterizations.FullPDFs import PdfConfiguration as PdfConfig
pdfConfig = PdfConfig()
# read parameters from file
print 120 * '='
pdfConfig.readParametersFromFile( filePath = args.parFilePathIn )
pdfConfig.setParametersInPdf(simMassPdf)
print 'P2VV - INFO: splitDataSetSigBkg: observables in mass PDF:'
massPdfObs.Print('v')
print 'P2VV - INFO: splitDataSetSigBkg: parameters in mass PDF:'
massPdfPars.Print('v')
###########################################################################################################################################
## split data into signal and background ##
###########################################
