def _print(c, p, e, l):
plot.seterror(e)
if not e:
e = 0x0
plot.set_ErrorBand(e)
c.Clear()
p.Draw(c,1,ratio,errband)
c.Modified()
c.Update()
hwwtools.ensuredir(opts.output)
outbasename = os.path.join(opts.output,'fitshapes_mH%d_%s_%s' % (opt.mass,bin,mode))
if l:
outbasename += '_' + l
c.Print(outbasename+'.pdf')
c.Print(outbasename+'.png')
def _print(c, p, e, l):
plot.seterror(e)
if not e:
e = 0x0
plot.set_ErrorBand(e)
c.Clear()
p.Draw(c, 1, ratio, errband)
c.Modified()
c.Update()
hwwtools.ensuredir(opts.output)
outbasename = os.path.join(
opts.output, 'fitshapes_mH%d_%s_%s' % (opt.mass, bin, mode))
if l:
outbasename += '_' + l
c.Print(outbasename + '.pdf')
c.Print(outbasename + '.png')
siname = 'hww-{lumi:.2f}fb.mH{mass}.{channel}_shape'
# masses = [m for m in hwwinfo.masses if m <= 250] if opt.mass == 0 else [opt.mass]
masses = [m for m in opt.mass if m <= 250]
dcdir = opt.input + '/datacards'
shdir = dcdir + '/shapes'
rndm = ROOT.TRandom3()
for k in xrange(opt.ntoys):
print '==> Instance', k
toypath = 'toys/instance_%d/' % k
toydcpath = toypath + 'datacards'
toyshpath = toypath + 'datacards/shapes'
hwwtools.ensuredir(toydcpath)
hwwtools.ensuredir(toydcpath + '/pseudo')
if os.path.exists(toyshpath):
os.unlink(toyshpath)
os.symlink(os.path.abspath(shdir), toyshpath)
for chan in opt.chans:
# the file for signal injection must be unique for all masses
# let's take 125 as a reference
refname = siname.format(lumi=opt.lumi, mass=125, channel=chan)
refpath = os.path.join(shdir, refname + '.root')
reffile = ROOT.TFile.Open(refpath)
data_si = reffile.Get('histo_Data')
if not data_si.__nonzero__():
reffile.Close()
def fitAndPlot( dcpath, opts ):
'''
1. read the datacard
2. convert to ws
3. run combine
4. open get the mlfit rootfile
1-4 don't need to know the content of the card, only to check that there are shapes inside
'''
remass = re.compile('mH(\d*)')
m = remass.search(dcpath)
if not m: raise ValueError('Mass not found! Name your datacards properly!')
print 'Mass',m.group(1)
opt.mass = int(m.group(1))
shapepath = os.path.join(os.getenv('CMSSW_BASE'),'src/HWWAnalysis/ShapeAnalysis')
print 'Shape directory is',shapepath
ROOT.gInterpreter.ExecuteMacro(shapepath+'/macros/LatinoStyle2.C')
# 1. load the datacard
dcfile = open(dcpath,'r')
class DCOptions: pass
options = DCOptions()
options.stat = False
options.bin = True
options.noJMax = False
options.nuisancesToExclude = []
options.nuisancesToRescale = []
options.fileName = dcpath
options.out = None
options.cexpr = False
options.fixpars = False
options.libs = []
options.verbose = 0
options.poisson = 0
options.mass = opt.mass
DC = parseCard(dcfile, options)
if not DC.hasShapes:
print 'This datacard has no shapes!'
print dcpath
sys.exit(-1)
# if len(DC.bins) != 1:
# raise ValueError('Only 1 bin datacards supported at the moment: '+', '.join(DC.bins))
# 2. convert to ws
wspath = os.path.splitext(dcpath)[0]+'_workspace.root'
logging.debug('Working with workspace %s',wspath)
print " wspath = ", wspath
mkws = (not os.path.exists(wspath) or
os.path.getmtime(wspath) < os.path.getmtime(dcpath) or
opts.clean)
print " mkws = ", mkws
print " not os.path.exists(wspath) = ", not os.path.exists(wspath)
if os.path.exists(wspath) :
print " os.path.getmtime(wspath) = ", os.path.getmtime(wspath)
print " os.path.getmtime(dcpath) = ", os.path.getmtime(dcpath)
print " opts.clean = ", opts.clean
if mkws:
# workspace + parameters = shapes
print 'Making the workspace...',
sys.stdout.flush()
if opts.combineFolder != None :
print " will run text2workspace in :", opts.combineFolder
curentDir = os.getcwd()
os.chdir( '%s' % (opts.combineFolder) )
#os.system( 'cd %s ' % (opts.combineFolder) )
#os.getenv('CMSSW_BASE')
print "( 'cd %s ' % (opts.combineFolder) ) = ", ( 'cd %s ' % (opts.combineFolder) )
print " pwd = "
os.system( 'pwd' )
#os.system( 'cmsenv' )
os.system( 'eval `scramv1 runtime -sh`' )
#os.system( 'cd - ' )
os.chdir( curentDir )
print " opts.modelName = ", opts.modelName
os.system( 'text2workspace.py %s -o %s %s' % (dcpath,wspath,opts.modelName) )
print 'done.'
ROOT.gSystem.Load('libHiggsAnalysisCombinedLimit')
wsfile = ROOT.TFile.Open(wspath)
if not wsfile.__nonzero__():
raise IOError('Could not open '+wspath)
w = wsfile.Get('w')
w.saveSnapshot('clean',w.allVars())
# run combine if requested
if opt.usefit:
mlfpath = opt.usefit
print '-'*80
print 'Using results in',mlfpath
print '-'*80
if not os.path.exists(mlfpath):
raise IOError('Fit result file %s not found.' % mlfpath )
else:
# 3.0 prepare the temp direcotry
import tempfile
mlfdir = opt.tmpdir if opt.tmpdir else tempfile.mkdtemp(prefix='mlfit_')
hwwtools.ensuredir(mlfdir)
# 3.1 go to the tmp dir
here = os.getcwd()
os.chdir(mlfdir)
# 3.2
mlcmd = ' '
if opts.injectionSignal == False :
mlcmd = 'combine -M MaxLikelihoodFit --saveNormalizations '+os.path.join(here,wspath)
else :
mlcmd = 'combine --expectSignal=1 -t -1 -M MaxLikelihoodFit --saveNormalizations '+os.path.join(here,wspath)
logging.debug(mlcmd)
print 'do: ',mlcmd
print 'Fitting the workspace...',
sys.stdout.flush()
if opts.fit: os.system(mlcmd)
os.chdir(here)
print 'done.'
# 3.3 set the max-like fit results path
mlfpath = os.path.join(mlfdir,'mlfit.root')
# 4. open the output and get the normalizations
mlffile = ROOT.TFile.Open(mlfpath)
if not mlffile.__nonzero__():
raise IOError('Could not open '+mlfpath)
model_s = w.pdf('model_s')
model_b = w.pdf('model_b')
res_s = mlffile.Get('fit_s')
res_b = mlffile.Get('fit_b')
sig_fit = ( model_s, res_s.floatParsFinal(), mlffile.Get('norm_fit_s'), )
bkg_fit = ( model_b, res_b.floatParsFinal(), mlffile.Get('norm_fit_b'), )
print 'List of bins found',', '.join(DC.bins)
bin = DC.bins[0]
modes = odict.OrderedDict([
('init',(model_s,res_s.floatParsInit(),None)), #(None, None, model_s)
('bkg' ,bkg_fit),
('sig' ,sig_fit),
])
# experimental
MB = ShapeBuilder(DC, options)
allshapes = {}
nuisancemap = {}
for mode,fit in modes.iteritems():
print 'Analysing model:',mode
logging.debug('Plotting %s', fit)
allshapes[mode] = {}
for bin in DC.bins:
print ' - Bin:',bin
coroner = Coroner(bin, DC, MB, w, fit)
coroner.errmode = opt.errmode
shapes,errs = coroner.perform()
nuisancemap[bin] = coroner.nuisances()
if opts.output:
printshapes(shapes, errs, mode, opts, bin, DC.signals, DC.processes)
allshapes[mode][bin] = (shapes,errs)
if opts.dump:
logging.debug('Dumping histograms to %s',opts.dump)
dumpdir = os.path.dirname(opt.dump)
# open rootfile
if dumpdir: hwwtools.ensuredir(dumpdir)
dump = ROOT.TFile.Open(opts.dump,'recreate')
here = ROOT.gDirectory.func()
dump.cd()
idir = dump.mkdir('info')
idir.cd()
# save the list of nuisances
nuisances = ROOT.TObjArray() #ROOT.std.vector('string')()
for (n,nf,pf,a,e) in DC.systs: nuisances.Add( ROOT.TObjString(n) )
nuisances.Write('nuisances', ROOT.TObject.kSingleKey)
# save the list of processes
processes = ROOT.TObjArray() #ROOT.std.vector('string')()
for p in DC.processes: processes.Add( ROOT.TObjString(p) )
processes.Write('processes', ROOT.TObject.kSingleKey)
# save the list of signals
signals = ROOT.TObjArray() #ROOT.std.vector('string')()
for s in DC.signals: signals.Add( ROOT.TObjString(s) )
signals.Write('signals', ROOT.TObject.kSingleKey)
# save the list of nuisances per bin
nuisbybin = ROOT.TMap()
for bin,nuis in nuisancemap.iteritems():
tnuis = ROOT.TObjArray()
for n in nuis: tnuis.Add( ROOT.TObjString(n) )
nuisbybin.Add(ROOT.TObjString(bin),tnuis)
nuisbybin.Write('map_binnuisances', ROOT.TObject.kSingleKey)
for mode,allbins in allshapes.iteritems():
# make the main directory
mdir = dump.mkdir(mode)
mdir.cd()
# info directory
idir = mdir.mkdir('info')
idir.cd()
# save the fit parameters
model,pars,norms = modes[mode]
pars.Write('parameters')
# save the list of signals
# save the bin plots
for bin,(shapes,errs) in allbins.iteritems():
# bin directory
bdir = mdir.mkdir(bin)
bdir.cd()
for s in shapes.itervalues():
# print s
s.Write()
for p,nugs in errs.iteritems():
dp = bdir.mkdir(p)
dp.cd()
for g in nugs.itervalues(): g.Write()
bdir.cd()
try:
modelall = errs['model']['all'].Clone('model_errs')
modelall.SetTitle('model_errs')
modelall.Write()
except:
logging.warn('Error graph model:err not found')
mdir.cd()
dump.Write()
dump.Close()
here.cd()
siname = 'hww-{lumi:.2f}fb.mH{mass}.{channel}_shape'
# masses = [m for m in hwwinfo.masses if m <= 250] if opt.mass == 0 else [opt.mass]
masses = [m for m in opt.mass if m <= 250]
dcdir = opt.input+'/datacards'
shdir = dcdir+'/shapes'
rndm = ROOT.TRandom3()
for k in xrange(opt.ntoys):
print '==> Instance',k
toypath = 'toys/instance_%d/' % k
toydcpath = toypath+'datacards'
toyshpath = toypath+'datacards/shapes'
hwwtools.ensuredir( toydcpath )
hwwtools.ensuredir( toydcpath+'/pseudo' )
if os.path.exists(toyshpath):
os.unlink(toyshpath)
os.symlink(os.path.abspath(shdir),toyshpath)
for chan in opt.chans:
# the file for signal injection must be unique for all masses
# let's take 125 as a reference
refname = siname.format(lumi=opt.lumi, mass=125, channel=chan)
refpath = os.path.join(shdir,refname+'.root')
reffile = ROOT.TFile.Open(refpath)
data_si = reffile.Get('histo_Data')
if not data_si.__nonzero__():
reffile.Close()
def fitAndPlot(dcpath, opts):
'''
1. read the datacard
2. convert to ws
3. run combine
4. open get the mlfit rootfile
1-4 don't need to know the content of the card, only to check that there are shapes inside
'''
remass = re.compile('mH(\d*)')
m = remass.search(dcpath)
if not m: raise ValueError('Mass not found! Name your datacards properly!')
print 'Mass', m.group(1)
opt.mass = int(m.group(1))
shapepath = os.path.join(os.getenv('CMSSW_BASE'),
'src/HWWAnalysis/ShapeAnalysis')
print 'Shape directory is', shapepath
ROOT.gInterpreter.ExecuteMacro(shapepath + '/macros/LatinoStyle2.C')
# 1. load the datacard
dcfile = open(dcpath, 'r')
class DCOptions:
pass
options = DCOptions()
options.stat = False
options.bin = True
options.noJMax = False
options.nuisancesToExclude = []
options.nuisancesToRescale = []
options.fileName = dcpath
options.out = None
options.cexpr = False
options.fixpars = False
options.libs = []
options.verbose = 0
options.poisson = 0
options.mass = opt.mass
DC = parseCard(dcfile, options)
if not DC.hasShapes:
print 'This datacard has no shapes!'
print dcpath
sys.exit(-1)
# if len(DC.bins) != 1:
# raise ValueError('Only 1 bin datacards supported at the moment: '+', '.join(DC.bins))
# 2. convert to ws
wspath = os.path.splitext(dcpath)[0] + '_workspace.root'
logging.debug('Working with workspace %s', wspath)
mkws = (not os.path.exists(wspath)
or os.path.getmtime(wspath) < os.path.getmtime(dcpath)
or opts.clean)
if mkws:
# workspace + parameters = shapes
print 'Making the workspace...',
sys.stdout.flush()
os.system('text2workspace.py %s -o %s' % (dcpath, wspath))
print 'done.'
ROOT.gSystem.Load('libHiggsAnalysisCombinedLimit')
wsfile = ROOT.TFile.Open(wspath)
if not wsfile.__nonzero__():
raise IOError('Could not open ' + wspath)
w = wsfile.Get('w')
w.saveSnapshot('clean', w.allVars())
# run combine if requested
if opt.usefit:
mlfpath = opt.usefit
print '-' * 80
print 'Using results in', mlfpath
print '-' * 80
if not os.path.exists(mlfpath):
raise IOError('Fit result file %s not found.' % mlfpath)
else:
# 3.0 prepare the temp direcotry
import tempfile
mlfdir = opt.tmpdir if opt.tmpdir else tempfile.mkdtemp(
prefix='mlfit_')
hwwtools.ensuredir(mlfdir)
# 3.1 go to the tmp dir
here = os.getcwd()
os.chdir(mlfdir)
# 3.2
mlcmd = 'combine -M MaxLikelihoodFit --saveNormalizations ' + os.path.join(
here, wspath)
logging.debug(mlcmd)
print 'Fitting the workspace...',
sys.stdout.flush()
if opts.fit: os.system(mlcmd)
os.chdir(here)
print 'done.'
# 3.3 set the max-like fit results path
mlfpath = os.path.join(mlfdir, 'mlfit.root')
# 4. open the output and get the normalizations
mlffile = ROOT.TFile.Open(mlfpath)
if not mlffile.__nonzero__():
raise IOError('Could not open ' + mlfpath)
model_s = w.pdf('model_s')
model_b = w.pdf('model_b')
res_s = mlffile.Get('fit_s')
res_b = mlffile.Get('fit_b')
sig_fit = (
model_s,
res_s.floatParsFinal(),
mlffile.Get('norm_fit_s'),
)
bkg_fit = (
model_b,
res_b.floatParsFinal(),
mlffile.Get('norm_fit_b'),
)
print 'List of bins found', ', '.join(DC.bins)
bin = DC.bins[0]
modes = odict.OrderedDict([
('init', (model_s, res_s.floatParsInit(),
None)), #(None, None, model_s)
('bkg', bkg_fit),
('sig', sig_fit),
])
# experimental
MB = ShapeBuilder(DC, options)
allshapes = {}
nuisancemap = {}
for mode, fit in modes.iteritems():
print 'Analysing model:', mode
logging.debug('Plotting %s', fit)
allshapes[mode] = {}
for bin in DC.bins:
print ' - Bin:', bin
coroner = Coroner(bin, DC, MB, w, fit)
coroner.errmode = opt.errmode
shapes, errs = coroner.perform()
nuisancemap[bin] = coroner.nuisances()
if opts.output:
printshapes(shapes, errs, mode, opts, bin, DC.signals,
DC.processes)
allshapes[mode][bin] = (shapes, errs)
if opts.dump:
logging.debug('Dumping histograms to %s', opts.dump)
dumpdir = os.path.dirname(opt.dump)
# open rootfile
if dumpdir: hwwtools.ensuredir(dumpdir)
dump = ROOT.TFile.Open(opts.dump, 'recreate')
here = ROOT.gDirectory.func()
dump.cd()
idir = dump.mkdir('info')
idir.cd()
# save the list of nuisances
nuisances = ROOT.TObjArray() #ROOT.std.vector('string')()
for (n, nf, pf, a, e) in DC.systs:
nuisances.Add(ROOT.TObjString(n))
nuisances.Write('nuisances', ROOT.TObject.kSingleKey)
# save the list of processes
processes = ROOT.TObjArray() #ROOT.std.vector('string')()
for p in DC.processes:
processes.Add(ROOT.TObjString(p))
processes.Write('processes', ROOT.TObject.kSingleKey)
# save the list of signals
signals = ROOT.TObjArray() #ROOT.std.vector('string')()
for s in DC.signals:
signals.Add(ROOT.TObjString(s))
signals.Write('signals', ROOT.TObject.kSingleKey)
# save the list of nuisances per bin
nuisbybin = ROOT.TMap()
for bin, nuis in nuisancemap.iteritems():
tnuis = ROOT.TObjArray()
for n in nuis:
tnuis.Add(ROOT.TObjString(n))
nuisbybin.Add(ROOT.TObjString(bin), tnuis)
nuisbybin.Write('map_binnuisances', ROOT.TObject.kSingleKey)
for mode, allbins in allshapes.iteritems():
# make the main directory
mdir = dump.mkdir(mode)
mdir.cd()
# info directory
idir = mdir.mkdir('info')
idir.cd()
# save the fit parameters
model, pars, norms = modes[mode]
pars.Write('parameters')
# save the list of signals
# save the bin plots
for bin, (shapes, errs) in allbins.iteritems():
# bin directory
bdir = mdir.mkdir(bin)
bdir.cd()
for s in shapes.itervalues():
# print s
s.Write()
for p, nugs in errs.iteritems():
dp = bdir.mkdir(p)
dp.cd()
for g in nugs.itervalues():
g.Write()
bdir.cd()
try:
modelall = errs['model']['all'].Clone('model_errs')
modelall.SetTitle('model_errs')
modelall.Write()
except:
logging.warn('Error graph model:err not found')
mdir.cd()
dump.Write()
dump.Close()
here.cd()
