def __call__(self, data):
pdf_pars = self.__pdf.getParameters(data.get())
if not self.__parameters:
self.__parameters = pdf_pars
self.__parameters = [p for p in pdf_pars] + [self.__status]
else:
for p in self.__parameters:
pdf_par = pdf_pars.find(p.GetName())
if not pdf_par:
continue
pdf_par.setVal(p.getVal())
pdf_par.setError(p.getError())
success = False
status = 4
for i in range(3):
self.__result = self.__pdf.fitTo(data, **self.__fit_opts)
status = self.__result.status()
if status == 0:
success = True
break
if status < 4:
self.__status.setIndex(self.__result.status())
else:
self.__status.setIndex(4)
if success:
from P2VV.Utilities.SWeights import SData
self.__sData = SData(Pdf=self.__pdf, Data=data, Name="MassSPlot")
data = self.__sData.data(self.__comp)
if self.__correct_weights:
from P2VV.Utilities.DataHandling import correctWeights
data = correctWeights(data, splitCatNames=None, ImportIntoWS=False)
return data
else:
return None
from ROOT import TCanvas
corr_canvas = TCanvas('corr_canvas', 'corr_canvas', 1000, 500)
corr_canvas.Divide(2, 1)
from ROOT import gStyle
gStyle.SetPalette(1)
## Plot correlation histogram
corr_canvas.cd(1)
corr_hist_sWmass = sWeight_mass_result.correlationHist()
corr_hist_sWmass.GetXaxis().SetLabelSize(0.03)
corr_hist_sWmass.GetYaxis().SetLabelSize(0.03)
corr_hist_sWmass.SetContour(20)
corr_hist_sWmass.Draw('colz')
from P2VV.Utilities.SWeights import SData
sData = SData(Pdf = sWeight_mass_pdf, Data = data, Name = 'SimulMassSPlot')
sig_sdata = sData.data('signal')
bkg_sdata = sData.data('background')
# Plot mass pdf
paper_calibration = FormulaVar(Name = 'st_calibration', Formula = '1.45 * @0', Arguments = [st])
dg_calibration = FormulaVar(Name = 'dg_calibration', Formula = "-6.81838 * @0 * @0 + 1.72 * @0", Arguments = [st])
sg_calibration = FormulaVar(Name = 'sg_calibration', Formula = "-0.00455 + 1.92 * @0-9.9352 * @0 * @0", Arguments = [st])
linear_calibration = FormulaVar(Name = 'linear_calibration', Formula = "0.00555 + 1.267 * @0", Arguments = [st])
from P2VV.Dilution import dilution
r = signal_dilution(psi_sdata, st, paper_calibration)
r = signal_dilution(psi_sdata, st, dg_calibration)
# r = signal_dilution(sig_sdata, st, sg_calibration)
# r = signal_dilution(sig_sdata, st, linear_calibration)
def __init__(self, time, masses, t_diff = None, MassResult = None,
InputFile = "/bfys/raaij/p2vv/data/Bs2JpsiPhiPrescaled_2011.root",
Workspace = 'Bs2JpsiPhiPrescaled_2011_workspace', Data = 'data',
UseKeysPdf = False, Weights = 'B', Draw = False, Reweigh = {}):
assert(Weights in ShapeBuilder.__weights)
self.__weights = Weights
self.__time = time
self.__t_diff = t_diff
pos = Workspace.find('201')
self.__year = Workspace[pos : pos + 4]
self.__input_ws = None
self.__ws = RooObject().ws()
self.__shapes = {}
self.__diff__shapes = {}
self.__masses = masses
self._sig = Component('wpv_signal', [], Yield = (1000, 10, 5e5))
self._psi = Component('wpv_jpsi', [], Yield = (5000, 10, 5e5))
self._bkg = Component('wpv_bkg', [], Yield = (5000, 10, 5e5))
if 'B' in masses:
## m_sig_mean = RealVar('wpv_m_sig_mean', Unit = 'MeV', Value = 5365, MinMax = (5363, 5372))
## m_sig_sigma = RealVar('wpv_m_sig_sigma', Unit = 'MeV', Value = 10, MinMax = (1, 20))
## from ROOT import RooGaussian as Gaussian
## self._sig_mass = Pdf(Name = 'wpv_sig_m', Type = Gaussian, Parameters = (masses['B'], m_sig_mean, m_sig_sigma ))
self._sig_mass = BMassPdf(masses['B'], Name = 'wpv_sig_mass', ParNamePrefix = "wpv",
AvSigParameterisation = True)
self._bkg_mass = BBkgPdf(masses['B'], Name = 'wpv_bkg_mass', ParNamePrefix = "wpv",
m_bkg_exp = dict(Name = 'm_bkg_exp', Value = -0.0017,
MinMax = (-0.01, -0.00001)))
self._sig[masses['B']] = self._sig_mass.pdf()
self._psi[masses['B']] = self._bkg_mass.pdf()
self._bkg[masses['B']] = self._bkg_mass.pdf()
if 'jpsi' in masses:
self._sig_mpsi = PsiMassPdf(masses['jpsi'], Name = 'wpv_sig_mpsi', ParNamePrefix = "wpv")
self._bkg_mpsi = PsiBkgPdf(masses['jpsi'], Name = 'wpv_bkg_mpsi', ParNamePrefix = "wpv")
self._sig[masses['jpsi']] = self._sig_mpsi.pdf()
self._psi[masses['jpsi']] = self._sig_mpsi.pdf()
self._bkg[masses['jpsi']] = self._bkg_mpsi.pdf()
self.__components = {'jpsi' : dict(jpsi = self._psi, bkg = self._bkg),
'B' : dict(B = self._sig, bkg = self._bkg),
'both' : dict(B = self._sig, jpsi = self._psi, bkg = self._bkg)}
self.__pdf = buildPdf(self.__components[Weights].values(), Observables = masses.values(),
Name = 'wpv_mass_pdf')
if MassResult:
## Use the provided mass result to set all the parameter values, only float the yields
pdf_params = self.__pdf.getParameters(RooArgSet(*masses.values()))
for p in MassResult.floatParsFinal():
## ignore yields
if any((p.GetName().startswith(n) for n in ['N_', 'mpsi_c'])):
continue
## Find pdf parameter, add "wpv_" prefix
pdf_p = pdf_params.find('wpv_' + p.GetName())
if pdf_p:
pdf_p.setVal(p.getVal())
pdf_p.setError(p.getError())
pdf_p.setConstant(True)
self.__pdf.Print("t")
from ROOT import TFile
input_file = TFile.Open(InputFile)
if not input_file or not input_file.IsOpen():
raise OSError
if Workspace:
self.__input_ws = input_file.Get(Workspace)
if not self.__input_ws:
print 'Cannot find workspace %s in mixing file.' % Workspace
raise RuntimeError
self._data = self.__input_ws.data(Data)
if not self._data:
print 'Cannot find data in workspace %s.' % Workspace
raise RuntimeError
else:
self._data = input_file.Get(Data)
if not self._data.get().find(time.GetName()) and "refit" in time.GetName():
from ROOT import RooFormulaVar, RooArgList
def __add_alias(name, obs):
obs_name = obs.GetName()[:-6]
do = self._data.get().find(obs_name)
rf = RooFormulaVar(name, name, "@0", RooArgList(do))
a = self._data.addColumn(rf)
a.setMin(obs.getMin())
a.setMax(obs.getMax())
return a
## Add refit observables
time = __add_alias("time_refit", time)
self.__time = time
self.__time.Print()
if t_diff:
t_diff = __add_alias("time_diff_refit", t_diff)
self.__t_diff = t_diff
if t_diff:
self._data = self._data.reduce("{0} > {1} && {0} < {2} && {3} > {4} && {3} < {5}".format(time.GetName(), time.getMin(), time.getMax(), t_diff.GetName(), t_diff.getMin(), t_diff.getMax()))
else:
self._data = self._data.reduce("{0} > {1} && {0} < {2}".format(time.GetName(), time.getMin(), time.getMax()))
# self._data = self._data.reduce("mass > 5348 && mass < 5388")
fitOpts = dict(NumCPU = 4, Save = True, Minimizer = 'Minuit2', Optimize = 2)
self.__result = self.__pdf.fitTo(self._data, **fitOpts)
from P2VV.Utilities.SWeights import SData
for p in self.__pdf.Parameters(): p.setConstant(not p.getAttribute('Yield'))
splot = SData(Pdf = self.__pdf, Data = self._data, Name = 'MixingMassSplot')
self.__sdatas = {}
self.__reweigh_weights = {}
for key, c in self.__components[Weights].iteritems():
sdata = splot.data(c.GetName())
if 'Data' in Reweigh and key in Reweigh['Data']:
from array import array
from ROOT import RooBinning
source = Reweigh['Data'][key]
binning = Reweigh['Binning']
if type(binning) == array:
binning = RooBinning(len(binning) - 1, binning)
binning.SetName('reweigh')
Reweigh['DataVar'].setBinning(binning, 'reweigh')
source_obs = source.get().find('nPV')
cat_name = 'nPVs_' + key
source_cat = source.get().find(cat_name)
if source_cat:
# Remove previous weights to make sure we get it right
new_vars = source.get()
new_vars.remove(source_cat)
source = source.reduce(new_vars)
source_obs = source.get().find(source_obs.GetName())
source_obs.setBinning(binning, 'reweigh')
source_cat = BinningCategory(Name = cat_name, Observable = source_obs,
Binning = binning, Data = source, Fundamental = True)
from P2VV.Reweighing import reweigh
sdata, weights = reweigh(sdata, sdata.get().find('nPV'),
source, source_cat)
self.__reweigh_weights[key] = weights
sdata = self.__ws.put(sdata)
self.__sdatas[c] = sdata
rho_keys = dict((v, k) for k, v in self.__components[Weights].iteritems())
self.__shapes = {}
self.__diff_shapes = {}
for c, sdata in self.__sdatas.iteritems():
if UseKeysPdf:
rk = rho_keys[c]
rho = ShapeBuilder.__rho[rk] if rk in ShapeBuilder.__rho else 1.
time_shape = KeysPdf(Name = 'wpv_%s_pdf' % c.GetName(), Observable = time, Data = sdata, Rho = rho)
if t_diff:
diff_shape = KeysPdf(Name = 'wpv_%s_diff_pdf' % c.GetName(), Observable = t_diff, Data = sdata)
else:
time_shape = HistPdf(Name = 'wpv_%s_pdf' % c.GetName(), Observables = [time],
Data = sdata, Binning = {time : 35})
if t_diff:
diff_shape = HistPdf(Name = 'wpv_%s_diff_pdf' % c.GetName(), Observables = [t_diff],
Data = sdata, Binning = {time : 35})
self.__shapes[c] = time_shape
if t_diff:
self.__diff_shapes[c] = diff_shape
if Draw:
self.__draw()
if parFileOut :
# write parameters to file
pdfConfig.getParametersFromPdf( sWeightMassPdf, preDataSet )
pdfConfig.writeParametersToFile( filePath = parFileOut )
###########################################################################################################################################
## compute sWeights ##
######################
print 120 * '='
print 'P2VV - INFO: createBs2JpsiKKDataSet: computing sWeights'
# compute sWeights
from P2VV.Utilities.SWeights import SData
sWeightsBuilder = SData( Pdf = sWeightMassPdf, Data = preDataSet, Name = 'JpsiKK' )
preSDataSet = sWeightsBuilder.data()
# import sWeight variables
weightVars = [ ]
for wName in sWeightNames :
ws.put( preSDataSet.get().find(wName) )
weightVars.append( RealVar(wName) )
###################################################################################################################################
## add tagging observables to data sets ##
##########################################
print 120 * '='
print 'P2VV - INFO: createBs2JpsiKKDataSet: building tagging categories'
fitOpts = dict(NumCPU = 4, Timer = 1, Save = True, Optimize = 2,
Strategy = 2, Offset = True, Minimizer = 'Minuit2')
# Acceptance data
from ROOT import TFile
f= TFile.Open('Bs2JpsiPhi_DataSet.root')
assert f
workspace = f.Get('Bs2JpsiPhi_workspace')
data = workspace.data('data')
pdf.fitTo(data, **fitOpts)
# sPlot
from P2VV.Utilities.SWeights import SData
sData = SData(Pdf = pdf, Data = data, Name = 'splot')
# Observables
observables = data.get()
# Mapping for Fit
ranges = [o for o in observables if o.GetName().find('tp') != -1 ]
datas = dict([(c.GetName(), sData.data(c.GetName())) for c in (signal, comb_background, psi_background)])
yields = dict([(c.GetName(), c.getYield()) for c in (signal, comb_background, psi_background)])
weights = dict([( n, observables.find(o.GetName() + '_sw') ) for n, o in yields.iteritems()])
ranges.sort()
range_names = []
for i in range(0, len(ranges), 2):
l = ranges[i]
r = ranges[i + 1]
class SWeightTransform(object):
def __init__(self, pdf, component, fit_opts, correct_weights=True):
self.__comp = component
self.__pdf = pdf
self.__fit_opts = fit_opts
self.__result = None
self.__correct_weights = correct_weights
from ROOT import RooCategory
self.__status = RooCategory("sweight_status", "sweight fit status")
self.__status.defineType("success", 0)
self.__status.defineType("one", 1)
self.__status.defineType("two", 2)
self.__status.defineType("three", 3)
self.__status.defineType("other", 4)
self.__parameters = None
def __call__(self, data):
pdf_pars = self.__pdf.getParameters(data.get())
if not self.__parameters:
self.__parameters = pdf_pars
self.__parameters = [p for p in pdf_pars] + [self.__status]
else:
for p in self.__parameters:
pdf_par = pdf_pars.find(p.GetName())
if not pdf_par:
continue
pdf_par.setVal(p.getVal())
pdf_par.setError(p.getError())
success = False
status = 4
for i in range(3):
self.__result = self.__pdf.fitTo(data, **self.__fit_opts)
status = self.__result.status()
if status == 0:
success = True
break
if status < 4:
self.__status.setIndex(self.__result.status())
else:
self.__status.setIndex(4)
if success:
from P2VV.Utilities.SWeights import SData
self.__sData = SData(Pdf=self.__pdf, Data=data, Name="MassSPlot")
data = self.__sData.data(self.__comp)
if self.__correct_weights:
from P2VV.Utilities.DataHandling import correctWeights
data = correctWeights(data, splitCatNames=None, ImportIntoWS=False)
return data
else:
return None
def gen_params(self, observables=None):
from ROOT import RooArgSet
if self.__parameters:
return self.__parameters
else:
if observables and not isinstance(observables, RooArgSet):
obs = RooArgSet()
for o in observables:
obs.add(o._target_() if hasattr(o, "_target_") else o)
observables = obs
params = self.__pdf.getParameters(observables)
self.__parameters = [p for p in params] + [self.__status]
return self.__parameters
def result_params(self):
if not self.__result:
return []
else:
return [p for p in self.__result.floatParsFinal()] + [self.__status]
def set_params(self, data_params):
from ROOT import RooCategory
for trans_param in self.result_params():
data_param = data_params.find(trans_param.GetName())
if isinstance(trans_param, RooCategory):
data_param.setIndex(trans_param.getIndex())
else:
data_param.setVal(trans_param.getVal())
# This sets a symmetric error, but since we don't run Minos, that's ok
data_param.setError(trans_param.getError())
xTitle = 'M_{#mu^{+}#mu^{-}} [MeV/c^{2}]',
yTitle = 'Candidates / (%3.2f MeV/c^{2})' % ((mass_obs.getMax() - mass_obs.getMin()) / 50),
yTitleOffset = 1 / 0.7)
plots[''].append(ps)
from P2VV.Utilities.Resolution import plot_dir
if signal_MC:
plot_name = 'prescaled_%s_B_mass.pdf'
else:
plot_name = 'prescaled_%s_Jpsi_mass.pdf'
mass_canvas.Print(os.path.join(plot_dir, plot_name % args[0]), EmbedFonts = True)
from P2VV.Utilities.SWeights import SData
from P2VV.Utilities.DataHandling import correctWeights
data_clone = data.Clone(data.GetName())
sData = SData(Pdf = mass_pdf, Data = data_clone, Name = 'MassSPlot')
## single_bin_sig_sdata = correctWeights(sData.data(signal_name), splitCatNames = None, ImportIntoWS = False)
## single_bin_bkg_sdata = correctWeights(sData.data('background'), splitCatNames = None, ImportIntoWS = False)
single_bin_sig_sdata = sData.data(signal_name)
single_bin_bkg_sdata = sData.data('background')
del sData
sdatas = {'sig_sdata' : single_bin_sig_sdata,
'bkg_sdata' : single_bin_bkg_sdata}
## Mass PDF is still connected to data_clone, redirect and delete data_clone
mass_pdf.recursiveRedirectServers(data.get())
del data_clone
gc.collect()
if fit_mass and options.simultaneous:
from P2VV.Utilities.General import getSplitPar
# categories for splitting the PDF
simMassFitResult.correlationMatrix().Print()
else :
massPdfPars = massPdf.getVariables()
sWeightMassPdf = massPdf
###########################################################################################################################################
## compute sWeights ##
######################
print 'P2VV - INFO: createB2CCFitNTuple: computing sWeights'
# compute sWeights
from P2VV.Utilities.SWeights import SData
sWeightsBuilder = SData( Pdf = sWeightMassPdf, Data = dataSets['pre'][0], Name = 'JpsiKK' )
dataSets['preS'] = ( sWeightsBuilder.data(), [ ] )
# import sWeight variables
weightVars = [ ]
for wName in [ 'N_sigMass_sw', 'N_cbkgMass_sw' ] :
ws.put( dataSets['preS'][0].get().find(wName) )
weightVars.append( RealVar(wName) )
###########################################################################################################################################
## create signal and background data sets and print event yields ##
###################################################################
print 120 * '='
print 'P2VV - INFO: createB2CCFitNTuple: creating data set'
sigYield.setMin(0.0)
sigYield.setMax(nEvBin)
bkgYield.setVal(bkgYield.getVal() * nEvBin / nEv)
bkgYield.setError(sqrt(bkgYield.getVal()))
bkgYield.setMin(0.0)
bkgYield.setMax(nEvBin)
# determine mass parameters in each sub-sample with a fit
sim_mass_result = sWeight_mass_pdf.fitTo(data, **fitOpts)
sweight_pdf = sWeight_mass_pdf
else:
sweight_pdf = mass_pdf
from P2VV.Utilities.SWeights import SData
sdata = SData(Pdf=sweight_pdf, Data=data, Name="mass_sdata")
sig_sdata = sdata.data("signal")
bkg_sdata = sdata.data("background")
from ROOT import TFile
f = TFile.Open(input_file, "update")
tree = f.Get("DecayTree")
from ROOT import addVertexErrors, addJpsiDLS
from ROOT import std
dss = std.list("RooDataSet*")()
dss.push_back(sig_sdata)
dss.push_back(bkg_sdata)
if options.dls:
sigYield.setMin(0.)
sigYield.setMax(nEvBin)
bkgYield.setVal( bkgYield.getVal() * nEvBin / nEv )
bkgYield.setError( sqrt( bkgYield.getVal() ) )
bkgYield.setMin(0.)
bkgYield.setMax(nEvBin)
# determine mass parameters in each sub-sample with a fit
sim_mass_result = sWeight_mass_pdf.fitTo(data, **fitOpts)
sweight_pdf = sWeight_mass_pdf
else:
sweight_pdf = mass_pdf
from P2VV.Utilities.SWeights import SData
from P2VV.Utilities.DataHandling import correctWeights
sdata = SData( Pdf = sweight_pdf, Data = data, Name = 'mass_sdata')
sig_sdata = correctWeights(sdata.data('signal'), splitCatNames = [excl_biased.GetName()],
ImportIntoWS = False)
bkg_sdata = correctWeights(sdata.data('background'), splitCatNames = [excl_biased.GetName()],
ImportIntoWS = False)
from array import array
weights = array('d')
from ROOT import TFile
f = TFile.Open(input_file, "read")
tree = f.Get("DecayTree")
from ROOT import gDirectory
tree.Draw(">>elist", cut, "entrylist")
pdfOpts = dict()
plot(p, o, pdf = mass_pdf, data = data
, dataOpts = dict(MarkerSize = 0.8, MarkerColor = kBlack)
, pdfOpts = dict(LineWidth = 2, **pdfOpts)
, plotResidHist = True
, components = { 'bkg_*' : dict( LineColor = kRed, LineStyle = kDashed ),
## 'psi_*' : dict( LineColor = kGreen, LineStyle = kDashed ),
'sig_*' : dict( LineColor = kBlue, LineStyle = kDashed )
}
)
# Do the sWeights
# make sweighted dataset. TODO: use mumu mass as well...
from P2VV.Utilities.SWeights import SData
for p in mass_pdf.Parameters() : p.setConstant( not p.getAttribute('Yield') )
splot = SData(Pdf = mass_pdf, Data = data, Name = 'MassSplot')
data = splot.data('signal')
## psi_sdata = splot.data('psi_background')
bkg_sdata = splot.data('background')
if 'MC' in args[0]:
import random
## Set more events to be unbiased, so we get some HLT2 exclusive biased
## sample.
new_data = RooDataSet("new_data", "new_data", data.get())
for i, obs in enumerate(data):
b2 = obs.find('hlt2_biased')
ub2 = obs.find('hlt2_unbiased')
eb2 = obs.find('hlt2_excl_biased')
if b2.getIndex() == 0:
pass