def runPseudoExperiments(wsfile,pefile,experimentTag,options):
prepend = '[runPseudoExperiments %s] '%experimentTag
#read the file with input distributions
inputDistsF = ROOT.TFile.Open(pefile, 'READ')
print prepend+'Reading PE input from %s with %s' % (pefile, experimentTag)
if not checkKeyInFile(experimentTag, inputDistsF, doraise=False):
print prepend+"ERROR: experiment tag %s not found in input file %s" %(experimentTag, pefile)
sys.exit(-1)
wsInputFile = ROOT.TFile.Open(wsfile, 'READ')
ws = wsInputFile.Get('w')
wsInputFile.Close()
print prepend+'Read workspace from %s' % wsfile
#readout calibration from a file
calibMap=None
if options.calib:
cachefile = open(options.calib,'r')
calibMap = pickle.load(cachefile)
cachefile.close()
print prepend+'Read calibration from %s'%options.calib
genMtop=172.5
try:
genMtop=float(experimentTag.rsplit('_', 1)[1].replace('v','.'))
except (IndexError, ValueError) as e:
print ("%sERROR: Could not extract generated mass value "
"from experiment tag: '%s'"%(prepend, experimentTag))
sys.exit(-1)
print prepend+'Generated top mass is %5.1f GeV'%genMtop
#prepare results summary
selTag=''
if len(options.selection)>0 : selTag='_%s'%options.selection
summary=PseudoExperimentResults(genMtop=genMtop,
outFileUrl=osp.join(options.outDir,
'%s%s_results.root'%(experimentTag,selTag)),
selection=options.selection)
#load the model parameters and set all to constant
ws.loadSnapshot("model_params")
allVars = ws.allVars()
varIter = allVars.createIterator()
var = varIter.Next()
varCtr=0
while var :
varName=var.GetName()
if not varName in ['mtop', 'SVLMass', 'mu']:
ws.var(varName).setConstant(True)
varCtr+=1
var = varIter.Next()
print prepend+'setting to constant %d numbers in the model'%varCtr
ws.var("mtop").setVal(genMtop)
ws.var("mu").setVal(1.0)
#build the relevant PDFs
print prepend+"Building pdfs"
channels = ['em','mm','ee','m','e','eplus','mplus','eminus','mminus']
allPdfs = buildPDFs(ws=ws, options=options,
channels=channels,
calibMap=calibMap,
prepend=prepend)
#throw pseudo-experiments
poi = ROOT.RooArgSet( ws.var('mtop') )
if options.verbose>1:
print prepend+'Running %d experiments' % options.nPexp
print 80*'-'
# if not 'nominal' in experimentTag:
# cfilepath = osp.abspath(osp.join(osp.dirname(wsfile),'../../'))
# # cfilepath = osp.abspath(osp.join(osp.dirname(__file__),'../'))
# cfilepath = osp.join(cfilepath, ".svlsysthistos.pck")
# cachefile = open(cfilepath, 'r')
# systhistos = pickle.load(cachefile)
# print prepend+'>>> Read systematics histograms from cache (.svlsysthistos.pck)'
# cachefile.close()
for i in xrange(0,options.nPexp):
#iterate over available categories to build the set of likelihoods to combine
indNLLs=[]
nllMap={}
allPseudoDataH=[]
allPseudoData=[]
if options.verbose>1 and options.verbose<=3:
printProgress(i, options.nPexp, prepend+' ')
for key in sorted(allPdfs):
chsel, trk = key
mukey=(chsel+'_mu',trk)
if options.verbose>3:
sys.stdout.write(prepend+'Exp %-3d (%-12s, %d):' % (i+1, chsel, trk))
sys.stdout.flush()
ws.var('mtop').setVal(172.5)
ws.var('mu').setVal(1.0)
#read histogram and generate random data
ihist = inputDistsF.Get('%s/SVLMass_%s_%s_%d'%(experimentTag,chsel,experimentTag,trk))
# Get number of events to be generated either:
# - From properly scaled input files for nominal mass variations
# to estimate the actual statistical error
# - From the number of generated MC events, to estimate statistical
# uncertainty of variation
# THIS SCREWS UP THE MASS EXTRACTION: WHY??
nevtsSeed = ihist.Integral()
# if not 'nominal' in experimentTag:
# try:
# nevtsSeed = systhistos[(chsel, experimentTag.replace('_172v5',''),
# 'tot' ,trk)].GetEntries() ## FIXME: GetEntries or Integral?
# except KeyError:
# print prepend+" >>> COULD NOT FIND SYSTHISTO FOR",chsel, experimentTag, trk
# print '[Generation] Will generate PEs with %6.1f events' % nevtsSeed
nevtsToGen = ROOT.gRandom.Poisson(nevtsSeed)
pseudoDataH,pseudoData=None,None
if options.genFromPDF:
obs = ROOT.RooArgSet(ws.var('SVLMass'))
pseudoData = allPdfs[key].generateBinned(obs, nevtsToGen)
else:
pseudoDataH = ihist.Clone('peh')
if options.nPexp>1:
pseudoDataH.Reset('ICE')
pseudoDataH.FillRandom(ihist, nevtsToGen)
else:
print 'Single pseudo-experiment won\'t be randomized'
pseudoData = ROOT.RooDataHist('PseudoData_%s_%s_%d'%(experimentTag,chsel,trk),
'PseudoData_%s_%s_%d'%(experimentTag,chsel,trk),
ROOT.RooArgList(ws.var('SVLMass')), pseudoDataH)
if options.verbose>3:
sys.stdout.write(' [generated pseudodata]')
sys.stdout.flush()
#create likelihood : store it in the appropriate categories for posterior combination
nll = allPdfs[key].createNLL(pseudoData, ROOT.RooFit.Extended())
indNLLs.append( nll )
chType=''
if chsel in ['em','mm','ee'] : chType='ll'
if chsel in ['e','m'] : chType='lj'
if chsel in ['eplus','mplus'] : chType='lplus'
if chsel in ['eminus','minus'] : chType='lminus'
nllMapKeys=[('comb%s'%chsel,0),('comb%s'%chType,trk),('comb%s'%chType,0)]
if chType!='lplus' and chType!='lminus':
nllMapKeys.insert(0,('comb',trk))
nllMapKeys.insert(0,('comb',0))
for nllMapKey in nllMapKeys:
if not (nllMapKey in nllMap):
nllMap[nllMapKey]=[]
if nllMapKey[0]=='comb' and nllMapKey[1]==0:
nllMap[nllMapKey].append( nll )
else:
nllMap[nllMapKey].append( nll )
if options.verbose>3:
sys.stdout.write(' [running Minuit]')
sys.stdout.flush()
minuit=ROOT.RooMinuit(nll)
minuit.setErrorLevel(0.5)
minuit.migrad()
minuit.hesse()
minuit.minos(poi)
#save fit results
summary.addFitResult(key=key,ws=ws)
#show, if required
selstring = options.selection if options.selection else 'inclusive'
if options.spy and i==0:
pll=nll.createProfile(poi)
showFinalFitResult(data=pseudoData,pdf=allPdfs[key], nll=[pll],
SVLMass=ws.var('SVLMass'),mtop=ws.var('mtop'),
outDir=options.outDir,
tag=[selstring,
"%s channel, =%s tracks"%(
str(chsel.split('_',1)[0]),
str(trk))])
#raw_input('press key to continue...')
#save to erase later
if pseudoDataH : allPseudoDataH.append(pseudoDataH)
allPseudoData.append(pseudoData)
if options.verbose>3:
sys.stdout.write('%s DONE %s'
'(mt: %6.2f+-%4.2f GeV, '
'mu: %4.2f+-%4.2f'%
(bcolors.OKGREEN, bcolors.ENDC,
ws.var('mtop').getVal(), ws.var('mtop').getError(),
ws.var('mu').getVal(), ws.var('mu').getError()))
if options.floatCorrFrac:
sys.stdout.write(' corfrac: %4.2f+-%4.2f)'% (
ws.var('ttcorfracshift_%s_%d'%(chsel,trk)).getVal()+1.0,
ws.var('ttcorfracshift_%s_%d'%(chsel,trk)).getError()))
sys.stdout.write('\n')
sys.stdout.flush()
#combined likelihoods
if options.verbose>3:
print '%s------ Combining channels and categories'%(prepend)
for key in sorted(nllMap.keys()):
#reset to central values
ws.var('mtop').setVal(172.5)
ws.var('mu').setVal(1.0)
#add the log likelihoods and minimize
llSet = ROOT.RooArgSet()
for ll in nllMap[key]: llSet.add(ll)
combll = ROOT.RooAddition("combll","combll",llSet)
minuit=ROOT.RooMinuit(combll)
minuit.setErrorLevel(0.5)
minuit.migrad()
minuit.hesse()
minuit.minos(poi)
summary.addFitResult(key=key,ws=ws)
combll.Delete()
if options.verbose>3:
try: catlabel = CATTOLABEL[('%s_%d'%key)]
except KeyError: catlabel = CATTOLABEL[('%s_%d'%key).replace('_%s'%options.selection, '')]
sys.stdout.write(prepend)
sys.stdout.write('%8s (%2d cats): ' % (catlabel,len(nllMap[key])))
resultstring = ('mt: %6.2f+-%4.2f GeV, mu: %5.3f+-%5.3f \n' % (
ws.var('mtop').getVal(), ws.var('mtop').getError(),
ws.var('mu').getVal(), ws.var('mu').getError()) )
if key == ('comb', 0):
resultstring = bcolors.BOLD + resultstring + bcolors.ENDC
sys.stdout.write(resultstring)
sys.stdout.flush()
print 80*'-'
#free used memory
for h in allPseudoDataH : h.Delete()
for d in allPseudoData : d.Delete()
for ll in indNLLs : ll.Delete()
summary.saveResults()
cfilepath = osp.abspath(osp.join(osp.dirname(wsfile),'../../'))
# cfilepath = osp.abspath(osp.join(osp.dirname(__file__),'../'))
cfilepath = osp.join(cfilepath, ".svlsysthistos.pck")
cachefile = open(cfilepath, 'r')
systhistos = pickle.load(cachefile)
print prepend+'>>> Read systematics histograms from cache (.svlsysthistos.pck)'
cachefile.close()
for i in xrange(0,options.nPexp):
#iterate over available categories to build the set of likelihoods to combine
nllMap={}
allPseudoDataH=[]
allPseudoData=[]
if options.verbose>1 and options.verbose<=3:
printProgress(i, options.nPexp, prepend+' ')
for key in sorted(allPdfs):
chsel, trk = key
mukey=(chsel+'_mu',trk)
if options.verbose>3:
sys.stdout.write(prepend+'Exp %-3d (%-2s, %d):' % (i+1, chsel, trk))
sys.stdout.flush()
ws.var('mtop').setVal(172.5)
ws.var('mu').setVal(1.0)
#read histogram and generate random data
ihist = inputDistsF.Get('%s/SVLMass_%s_%s_%d'%(experimentTag,chsel,experimentTag,trk))
# Get number of events to be generated either:
histos[model]['Btype']=ROOT.TH1F('Btype',';B-hadron type;B-hadrons',4,0,4)
histos[model]['Btype'].GetXaxis().SetBinLabel(1,'B^{0}')
histos[model]['Btype'].GetXaxis().SetBinLabel(2,'B^{#pm}')
histos[model]['Btype'].GetXaxis().SetBinLabel(3,'B_{s}')
histos[model]['Btype'].GetXaxis().SetBinLabel(4,'Others')
histos[model]['Btype'].SetTitle('%s'%model)
histos[model]['Btype'].Sumw2()
histos[model]['Btype'].SetDirectory(0)
histos[model]['Btype_wgt']=histos[model]['Btype'].Clone('Btype_wgt')
histos[model]['Btype_wgt'].SetDirectory(0)
totalEntries=int(0.2*tree.GetEntriesFast())
for i in xrange(0,totalEntries):
tree.GetEntry(i)
printProgress(i,totalEntries)
for ib in xrange(0,tree.nB):
bid,bidCtr='Others',3
if ROOT.TMath.Abs(tree.Bid[ib])==511 : bid,bidCtr='B0',0
elif ROOT.TMath.Abs(tree.Bid[ib])==521 : bid,bidCtr='Bpm',1
elif ROOT.TMath.Abs(tree.Bid[ib])==531 : bid,bidCtr='Bs',2
ptratio = tree.Bpt[ib]/tree.Bjpt[ib]
if ptratio>fragWgt[model][bid].GetXaxis().GetXmax() : continue
bwgt=fragWgt[model][bid].GetBinContent( fragWgt[model][bid].GetXaxis().FindBin(ptratio) )
histos[model]['Btype'].Fill(bidCtr)
histos[model]['Btype_wgt'].Fill(bidCtr,bwgt)
for bkey in ['inc',bid]:
def runPseudoExperiments(wsfile,pefile,experimentTag,options):
prepend = '[runPseudoExperiments %s] '%experimentTag
#read the file with input distributions
inputDistsF = ROOT.TFile.Open(pefile, 'READ')
print prepend+'Reading PE input from %s with %s' % (pefile, experimentTag)
if not checkKeyInFile(experimentTag, inputDistsF, doraise=False):
print prepend+"ERROR: experiment tag %s not found in input file %s" %(experimentTag, pefile)
sys.exit(-1)
wsInputFile = ROOT.TFile.Open(wsfile, 'READ')
ws = wsInputFile.Get('w')
wsInputFile.Close()
print prepend+'Read workspace from %s' % wsfile
#readout calibration from a file
calibMap=None
if options.calib:
cachefile = open(options.calib,'r')
calibMap = pickle.load(cachefile)
cachefile.close()
print prepend+'Read calibration from %s'%options.calib
genMtop=172.5
try:
genMtop=float(experimentTag.rsplit('_', 1)[1].replace('v','.'))
except (IndexError, ValueError) as e:
print ("%sERROR: Could not extract generated mass value "
"from experiment tag: '%s'"%(prepend, experimentTag))
sys.exit(-1)
print prepend+'Generated top mass is %5.1f GeV'%genMtop
#prepare results summary
selTag=''
if len(options.selection)>0 : selTag='_%s'%options.selection
summary=PseudoExperimentResults(genMtop=genMtop,
outFileUrl=osp.join(options.outDir,
'%s%s_results.root'%(experimentTag,selTag)),
selection=options.selection)
#load the model parameters and set all to constant
ws.loadSnapshot("model_params")
allVars = ws.allVars()
varIter = allVars.createIterator()
var = varIter.Next()
varCtr=0
while var :
varName=var.GetName()
if not varName in ['mtop', 'SVLMass', 'mu']:
ws.var(varName).setConstant(True)
varCtr+=1
var = varIter.Next()
print prepend+'setting to constant %d numbers in the model'%varCtr
ws.var("mtop").setVal(genMtop)
ws.var("mu").setVal(1.0)
#build the relevant PDFs
print prepend+"Building pdfs"
allPdfs = buildPDFs(ws=ws, options=options,
calibMap=calibMap,
prepend=prepend)
#throw pseudo-experiments
poi = ROOT.RooArgSet( ws.var('mtop') )
if options.verbose>1:
print prepend+'Running %d experiments' % options.nPexp
print 80*'-'
# if not 'nominal' in experimentTag:
# cfilepath = osp.abspath(osp.join(osp.dirname(wsfile),'../../'))
# # cfilepath = osp.abspath(osp.join(osp.dirname(__file__),'../'))
# cfilepath = osp.join(cfilepath, ".svlsysthistos.pck")
# cachefile = open(cfilepath, 'r')
# systhistos = pickle.load(cachefile)
# print prepend+'>>> Read systematics histograms from cache (.svlsysthistos.pck)'
# cachefile.close()
for i in xrange(0,options.nPexp):
#iterate over available categories to build the set of likelihoods to combine
indNLLs=[]
nllMap={}
allPseudoDataH=[]
allPseudoData=[]
if options.verbose>1 and options.verbose<=3:
printProgress(i, options.nPexp, prepend+' ')
for key in sorted(allPdfs):
chsel, trk = key
mukey=(chsel+'_mu',trk)
if options.verbose>3:
sys.stdout.write(prepend+'Exp %-3d (%-12s, %d):' % (i+1, chsel, trk))
sys.stdout.flush()
ws.var('mtop').setVal(172.5)
ws.var('mu').setVal(1.0)
#read histogram and generate random data
ihist = inputDistsF.Get('%s/SVLMass_%s_%s_%d'%(experimentTag,chsel,experimentTag,trk))
# Get number of events to be generated either:
# - From properly scaled input files for nominal mass variations
# to estimate the actual statistical error
# - From the number of generated MC events, to estimate statistical
# uncertainty of variation
# THIS SCREWS UP THE MASS EXTRACTION: WHY??
nevtsSeed = ihist.Integral()
# if not 'nominal' in experimentTag:
# try:
# nevtsSeed = systhistos[(chsel, experimentTag.replace('_172v5',''),
# 'tot' ,trk)].GetEntries() ## FIXME: GetEntries or Integral?
# except KeyError:
# print prepend+" >>> COULD NOT FIND SYSTHISTO FOR",chsel, experimentTag, trk
# print '[Generation] Will generate PEs with %6.1f events' % nevtsSeed
nevtsToGen = ROOT.gRandom.Poisson(nevtsSeed)
pseudoDataH,pseudoData=None,None
if options.genFromPDF:
obs = ROOT.RooArgSet(ws.var('SVLMass'))
pseudoData = allPdfs[key].generateBinned(obs, nevtsToGen)
else:
pseudoDataH = ihist.Clone('peh')
if options.nPexp>1:
pseudoDataH.Reset('ICE')
pseudoDataH.FillRandom(ihist, nevtsToGen)
else:
print 'Single pseudo-experiment won\'t be randomized'
pseudoData = ROOT.RooDataHist('PseudoData_%s_%s_%d'%(experimentTag,chsel,trk),
'PseudoData_%s_%s_%d'%(experimentTag,chsel,trk),
ROOT.RooArgList(ws.var('SVLMass')), pseudoDataH)
if options.verbose>3:
sys.stdout.write(' [generated pseudodata]')
sys.stdout.flush()
#create likelihood : store it in the appropriate categories for posterior combination
nll = allPdfs[key].createNLL(pseudoData, ROOT.RooFit.Extended())
indNLLs.append( nll )
chType=''
if chsel in ['em','mm','ee'] : chType='ll'
if chsel in ['e','m'] : chType='lj'
if chsel in ['eplus','mplus'] : chType='lplus'
if chsel in ['eminus','minus'] : chType='lminus'
nllMapKeys=[('comb%s'%chsel,0),('comb%s'%chType,trk),('comb%s'%chType,0)]
if chType!='lplus' and chType!='lminus':
nllMapKeys.insert(0,('comb',trk))
nllMapKeys.insert(0,('comb',0))
for nllMapKey in nllMapKeys:
if not (nllMapKey in nllMap):
nllMap[nllMapKey]=[]
if nllMapKey[0]=='comb' and nllMapKey[1]==0:
nllMap[nllMapKey].append( nll )
else:
nllMap[nllMapKey].append( nll )
if options.verbose>3:
sys.stdout.write(' [running Minuit]')
sys.stdout.flush()
minuit=ROOT.RooMinuit(nll)
minuit.setErrorLevel(0.5)
minuit.migrad()
minuit.hesse()
minuit.minos(poi)
#save fit results
summary.addFitResult(key=key,ws=ws)
#show, if required
selstring = options.selection if options.selection else 'inclusive'
if options.spy and i==0:
pll=nll.createProfile(poi)
showFinalFitResult(data=pseudoData,pdf=allPdfs[key], nll=[pll],
SVLMass=ws.var('SVLMass'),mtop=ws.var('mtop'),
outDir=options.outDir,
tag=[selstring,
"%s channel, =%s tracks"%(
str(chsel.split('_',1)[0]),
str(trk))])
#raw_input('press key to continue...')
#save to erase later
if pseudoDataH : allPseudoDataH.append(pseudoDataH)
allPseudoData.append(pseudoData)
if options.verbose>3:
sys.stdout.write('%s DONE %s'
'(mt: %6.2f+-%4.2f GeV, '
'mu: %4.2f+-%4.2f'%
(bcolors.OKGREEN, bcolors.ENDC,
ws.var('mtop').getVal(), ws.var('mtop').getError(),
ws.var('mu').getVal(), ws.var('mu').getError()))
if options.floatCorrFrac:
sys.stdout.write(' corfrac: %4.2f+-%4.2f)'% (
ws.var('ttcorfracshift_%s_%d'%(chsel,trk)).getVal()+1.0,
ws.var('ttcorfracshift_%s_%d'%(chsel,trk)).getError()))
sys.stdout.write('\n')
sys.stdout.flush()
#combined likelihoods
if options.verbose>3:
print '%s------ Combining channels and categories'%(prepend)
for key in sorted(nllMap.keys()):
#reset to central values
ws.var('mtop').setVal(172.5)
ws.var('mu').setVal(1.0)
#add the log likelihoods and minimize
llSet = ROOT.RooArgSet()
for ll in nllMap[key]: llSet.add(ll)
combll = ROOT.RooAddition("combll","combll",llSet)
minuit=ROOT.RooMinuit(combll)
minuit.setErrorLevel(0.5)
minuit.migrad()
minuit.hesse()
minuit.minos(poi)
summary.addFitResult(key=key,ws=ws)
combll.Delete()
if options.verbose>3:
try: catlabel = CATTOLABEL[('%s_%d'%key)]
except KeyError: catlabel = CATTOLABEL[('%s_%d'%key).replace('_%s'%options.selection, '')]
sys.stdout.write(prepend)
sys.stdout.write('%8s (%2d cats): ' % (catlabel,len(nllMap[key])))
resultstring = ('mt: %6.2f+-%4.2f GeV, mu: %5.3f+-%5.3f \n' % (
ws.var('mtop').getVal(), ws.var('mtop').getError(),
ws.var('mu').getVal(), ws.var('mu').getError()) )
if key == ('comb', 0):
resultstring = bcolors.BOLD + resultstring + bcolors.ENDC
sys.stdout.write(resultstring)
sys.stdout.flush()
print 80*'-'
#free used memory
for h in allPseudoDataH : h.Delete()
for d in allPseudoData : d.Delete()
for ll in indNLLs : ll.Delete()
summary.saveResults()
histos[model]['Btype'] = ROOT.TH1F('Btype', ';B-hadron type;B-hadrons', 4,
0, 4)
histos[model]['Btype'].GetXaxis().SetBinLabel(1, 'B^{0}')
histos[model]['Btype'].GetXaxis().SetBinLabel(2, 'B^{#pm}')
histos[model]['Btype'].GetXaxis().SetBinLabel(3, 'B_{s}')
histos[model]['Btype'].GetXaxis().SetBinLabel(4, 'Others')
histos[model]['Btype'].SetTitle('%s' % model)
histos[model]['Btype'].Sumw2()
histos[model]['Btype'].SetDirectory(0)
histos[model]['Btype_wgt'] = histos[model]['Btype'].Clone('Btype_wgt')
histos[model]['Btype_wgt'].SetDirectory(0)
totalEntries = tree.GetEntriesFast()
for i in xrange(0, totalEntries):
tree.GetEntry(i)
printProgress(i, totalEntries)
for ib in xrange(0, tree.nB):
bid, bidCtr = 'Others', 3
if ROOT.TMath.Abs(tree.Bid[ib]) == 511: bid, bidCtr = 'B0', 0
elif ROOT.TMath.Abs(tree.Bid[ib]) == 521: bid, bidCtr = 'Bpm', 1
elif ROOT.TMath.Abs(tree.Bid[ib]) == 531: bid, bidCtr = 'Bs', 2
ptratio = tree.Bpt[ib] / tree.Bjpt[ib]
if ptratio > fragWgt[model][bid].GetXaxis().GetXmax(): continue
bwgt = fragWgt[model][bid].GetBinContent(
fragWgt[model][bid].GetXaxis().FindBin(ptratio))
histos[model]['Btype'].Fill(bidCtr)
histos[model]['Btype_wgt'].Fill(bidCtr, bwgt)
