class ShapeBuilder(object):
__weights = set(('jpsi', 'B', 'both'))
__rho = dict(B = 2., jpsi = 2.)
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()
def __draw(self):
from ROOT import kDashed, kRed, kGreen, kBlue, kBlack
from ROOT import TCanvas
from ROOT import TPaveText
self.__year_label = TPaveText(0.71, 0.72, 0.89, 0.85, "NDC")
self.__year_label.SetFillColor(0)
self.__year_label.AddText(self.__year)
self.__year_label.SetBorderSize(0)
obs = self.__masses.values()
self.__canvas = TCanvas('wpv_canvas', 'wpv_canvas', len(obs) * 600, 533)
for (p,o) in zip(self.__canvas.pads(len(obs)), obs):
from P2VV.Utilities.Plotting import plot
plot(p, o, pdf = self.__pdf, data = self._data
, dataOpts = dict(MarkerSize = 0.8, MarkerColor = kBlack, Binning = 50)
, pdfOpts = dict(LineWidth = 2)
, plotResidHist = 'BX'
, xTitle = 'M_{#mu^{+}#mu^{-}} [MeV/c^{2}]'
, yTitle = 'Candidates / (%2.1f MeV/c^{2})' % ((o.getMax() - o.getMin()) / float(50))
, components = { 'wpv_bkg_*' : dict( LineColor = kRed, LineStyle = kDashed )
, 'wpv_psi_*' : dict( LineColor = kGreen, LineStyle = kDashed )
, 'wpv_sig_*' : dict( LineColor = kBlue, LineStyle = kDashed )
}
)
self.__canvas.cd(1)
self.__year_label.Draw()
self.__canvas.Update()
self.__draw_time()
if self.__t_diff:
self.__draw_t_diff()
def __draw_time(self):
from ROOT import kDashed, kRed, kGreen, kBlue, kBlack
from ROOT import TCanvas
from itertools import product
self.__time_canvases = []
t = self.__time
from P2VV.Utilities.Plotting import plot
from ROOT import RooArgSet
for logy, (c, shape) in product((True, False), self.__shapes.items()):
cn = 'wpv_time_%s' % c.GetName()
if logy: cn += '_log'
p = TCanvas(cn, cn, 600, 400)
self.__time_canvases.append(p)
nBins = 80
frame = plot(p, t, pdf = shape, data = self.__sdatas[c]
, frameOpts = dict(Title = c.GetName())
, dataOpts = dict(MarkerSize = 0.8, Binning = nBins, MarkerColor = kBlack)
, pdfOpts = dict(LineWidth = 2)
, logy = logy, yScale = (1 if logy else 0, None)
, plotResidHist = False)[0]
frame.GetXaxis().SetTitle('t [ps]')
if logy:
frame.GetYaxis().SetRangeUser(1, 7000 if self.__year == '2012' else 5000)
frame.GetYaxis().SetTitle('Candidates / (%3.2f ps)' % ((t.getMax() - t.getMin()) / float(nBins)))
self.__year_label.Draw()
from P2VV.Utilities.Resolution import plot_dir
p.Print(os.path.join(plot_dir, 'wpv_time_%s_%s_%s.pdf' % (c.GetName()[4:], self.__year, 'log' if logy else 'linear')), EmbedFonts = True)
def __draw_t_diff(self):
from ROOT import kDashed, kRed, kGreen, kBlue, kBlack
from ROOT import TCanvas
self.__diff_canvases = []
t_diff = self.__t_diff
from ROOT import RooArgSet
for c, shape in self.__diff_shapes.iteritems():
name = 'wpv_diff_%s' % c.GetName()
canvas = TCanvas(name, name, 600, 400)
p = canvas.cd()
self.__diff_canvases.append(canvas)
from P2VV.Utilities.Plotting import plot
plot(p, t_diff, pdf = shape, data = self.__sdatas[c]
, dataOpts = dict(MarkerSize = 0.8, Binning = 80, MarkerColor = kBlack)
, pdfOpts = dict(LineWidth = 2)
, xTitle = 't_{rec} - t_{true} [ps]'
, yTitle = 'Candidates / (12.5 fs)'
, logy = False
, plotResidHist = False)
self.__year_label.Draw()
from P2VV.Utilities.Resolution import plot_dir
canvas.Print(os.path.join(plot_dir, 'wpv_tdiff_%s_%s_linear.pdf' % (c.GetName()[4:], self.__year)), EmbedFonts = True)
def sdata(self, key):
c = self.__components[self.__weights][key]
return self.__sdatas[c]
def shape(self, key):
c = self.__components[self.__weights][key]
return self.__shapes[c]
def diff_shape(self, key):
c = self.__components[self.__weights][key]
return self.__diff_shapes[c]
def input_ws(self):
return self.__input_ws
def reweigh_weights(self, key):
try:
return self.__reweigh_weights[key]
except AttributeError:
return {}
dataFile.Write( dataFilePath, TObject.kOverwrite )
dataFile.Close()
else :
dataFile = TFile.Open(dataFilePath)
dataSets = dict( [ ( key, dataFile.Get( 'data_' + key ) ) for key in cutKeys ] )
dataFile.Close()
print 'build mass PDF'
from P2VV.RooFitWrappers import Component
sigMassComps = Component( 'sigMass', [ ], Yield = ( 85.e3, 0., 2.e5 ) )
bkgMassComps = Component( 'bkgMass', [ ], Yield = ( 1.e5, 0., 2.e5 ) )
yields = [ comp.getYield() for comp in [ sigMassComps, bkgMassComps ] ]
from P2VV.Parameterizations.MassPDFs import DoubleGauss_Signal_Mass as SignalBMass
sigMass = SignalBMass( mass = mass )
sigMassComps += sigMass.pdf()
from P2VV.Parameterizations.MassPDFs import LP2011_Background_Mass as BackgroundBMass
bkgMass = BackgroundBMass( mass = mass )
bkgMassComps += bkgMass.pdf()
from P2VV.RooFitWrappers import buildPdf
massPdf = buildPdf( [ sigMassComps, bkgMassComps ], Observables = [ mass ], Name = 'JpsiKKMass' )
ws['N_sigMass'].setVal(123.e3)
ws['m_sig_mean'].setVal(5368.2)
ws['m_sig_sigma_1'].setVal(5.6)
ws['m_sig_sigma_2'].setVal(14.)
ws['m_sig_frac'].setVal(0.65)
ws['m_bkg_exp'].setVal(-0.0015)
massPdf.getVariables().Print('v')
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()
selected = Category('sel', States = {'Selected' : 1, 'NotSelected' : 0})
observables = [t, m, mpsi, mphi, st, excl_biased, selected, zerr,
hlt1_biased, hlt1_unbiased, hlt2_biased, hlt2_unbiased] + angles
project_vars = [st, excl_biased]
# now build the actual signal PDF...
from ROOT import RooGaussian as Gaussian
from ROOT import RooExponential as Exponential
from ROOT import RooDecay as Decay
# B mass pdf
from P2VV.Parameterizations.MassPDFs import DoubleGauss_Signal_Mass as MassPdf
sig_m = MassPdf(m, Name = 'sig_m', AvSigParameterisation = True)
sig_m = sig_m.pdf()
# background
psi_c = RealVar( 'psi_c', Unit = '1/MeV', Value = -0.0004, MinMax = (-0.1, -0.0000001))
bkg_m = Pdf(Name = 'bkg_m', Type = Exponential, Parameters = [m, psi_c])
background = Component('background', (bkg_m,), Yield= (200000,500,5000000) )
signal = Component('signal', (sig_m,), Yield= (200000,500,500000) )
from P2VV.Utilities.DataHandling import readData
tree_name = 'DecayTree'
prefix = '/stuff/PhD' if os.path.exists('/stuff') else '/bfys/raaij'
if dataSample == '2011':
## input_file = '/stuff/PhD/p2vv/data/Bs2JpsiPhi_ntupleB_for_fitting_20121012_MagDownMagUp.root'
input_file = os.path.join(prefix, 'p2vv/data/Bs2JpsiPhiPrescaled_2012_ntupleB_20130722.root')
