def _calculate_confidence_band(root_obj: TF1,
ci: float,
steps: int = 1000) -> dict:
"""
Calculate confidence bands from a fit of a
TF1 to a histogram.
Parameters
----------
root_obj : TH1
ROOT TF1 pdf used for the fit
ci : float
Confidence interval
steps : int
Number of points in band
Returns
-------
dict
Dictionary with x, y, xerr, yerr points
"""
assert(isinstance(root_obj, TF1))
band = TGraphErrors(steps)
lowb, highb = root_obj.GetXaxis().GetXmin(), root_obj.GetXaxis().GetXmax()
for i in range(steps):
x = (lowb + (
(highb - lowb) / steps) * i)
y = 0
band.SetPoint(i, x, y)
(TVirtualFitter.GetFitter()).GetConfidenceIntervals(band, ci)
return _convert_tgrapherrors(band)
def TF2_Fit_NKG(h2,N0,r0,s0,x0,y0,x_min,x_max,y_min,y_max):
fitf2= TF2('fitf2',func_nkg_show,x_min,x_max,y_min,y_max,5)
fitf2.SetParameter(0,N0) # Setting starting values (Ne)
fitf2.SetParameter(1,r0) # ,, ,, (r_M)
fitf2.SetParameter(2,s0) # ,, ,, (s)
fitf2.SetParameter(3,x0) # ,, ,, (x_core)
fitf2.SetParameter(4,y0) # ,, ,, (y_core)
#//fitf2->SetParLimits(1,10,300) ; //Setting r_M limits
fitf2.SetParLimits(1,r0,r0)# //Setting r_M limits
#//fitf2->SetParLimits(2,1,2.5) ; //Setting Age limits (keeping fixed)
fitf2.SetParLimits(2,s0,s0)# //Setting Age limits (Keeping fixed)
h2.Fit(fitf2,"Q0")
#h2SetXTitle("X (m)") ;
#h2->GetXaxis()->CenterTitle() ;
#h2->SetYTitle("Y (m)") ;
#h2->GetYaxis()->CenterTitle() ;
#//h2->Draw("HIST") ;
#//fitf2->DrawCopy("surf1 same");
#gStyle->SetOptFit(1111) ;
Ne_fit=fitf2.GetParameter(0)
rM_fit=fitf2.GetParameter(1)
s_fit=fitf2.GetParameter(2)
x_core_fit=fitf2.GetParameter(3)
y_core_fit=fitf2.GetParameter(4)
x_core_fit_err=fitf2.GetParError(3)
y_core_fit_err=fitf2.GetParError(4)
fitter=TVirtualFitter.GetFitter()
corMatrix=np.zeros([5,5])
sig_i=0
sig_j=0
cov_ij=0
for i in np.arange(4):
sig_i=np.sqrt(fitter.GetCovarianceMatrixElement(i,i))
for j in np.arange(4):
sig_j=np.sqrt(fitter.GetCovarianceMatrixElement(j,j))
cov_ij=fitter.GetCovarianceMatrixElement(i,j)
corMatrix[i][j]=cov_ij/(sig_i*sig_j)
if(sig_i==0 or sig_j==0):
corMatrix[i][j]=0
corCoef_xy=corMatrix[1][2] #Correlation coeff. between the x & y core position
return Ne_fit,rM_fit,s_fit,x_core_fit,y_core_fit,x_core_fit_err,y_core_fit_err, corCoef_xy
def fit(self, graph, xstart, xend, options=''):
"""computes fit and stores calculated parameters with errors in parameter dict.
Also extracts covariance matrix and calculates correlation matrix
Arguments:
graph -- an instance of ROOTs TGraph with data to fit
xstart -- start value for x interval
xend -- end value for x interval
options -- fitting options (see ROOT documentation)
"""
graph.Fit(self._function, 'Q' + options, '', xstart,
xend) # Q = quit mode, no print out on console
# get fitted params
self.virtualFitter = TVirtualFitter.GetFitter()
if self.virtualFitter:
for i in self.params:
self.params[i]['value'] = self.virtualFitter.GetParameter(i)
self.params[i]['error'] = self.virtualFitter.GetParError(i)
# get cov. and corr. matrix
freeparams = dict(
) # map cov. matrix index to params index, only not fixed params are in cov. matrix
freecount = 0
fixedcount = 0
for i, param in self.params.items():
if param['fixed']:
fixedcount += 1
else:
freecount += 1
freeparams[i - fixedcount] = i
self._covMatrix = [[
self.virtualFitter.GetCovarianceMatrixElement(col, row)
for row in range(freecount)
] for col in range(freecount)]
self._corrMatrix = [[
self._covMatrix[col][row] /
(self.params[freeparams[col]]['error'] *
self.params[freeparams[row]]['error'])
for row in range(freecount)
] for col in range(freecount)]
logger.info("Do fitting")
logger.info(
"Argument \"setinitialgaussianmean\" set to %u, however not used at the moment",
setinitialgaussianmean)
if "Lb" not in case and "Lc" not in case and "D0" not in case and "Ds" not in case:
logger.warning(
"Can only do the fit for Lc, D0 or Ds, however found case %s",
case)
return -1, -1
histo.GetXaxis().SetTitle("Invariant Mass L_{c}^{+}(GeV/c^{2})")
histo.Rebin(rebin)
histo.SetStats(0)
TVirtualFitter.SetDefaultFitter("Minuit")
# Prepare what can be prepared
npar_sig = 3
npar_bkg = 3 if bkgfunc == "Pol2" else 2
par_names_sig = ["IntSig", "Mean", "Sigma"]
par_names_bkg = ["IntBkg", "BkgCoeff1", "BkgCoeff2"] if bkgfunc == "Pol2" \
else ["IntBkg", "BkgCoeff1"]
fitOption = "L,E" if dolikelihood else ""
logger.info("Initial parameters for signal fit are")
print(f"mean = {masspeak}\nsigma = {sigma_sig}")
logger.debug("fit background (just side bands)")
nSigma4SideBands = 4.
def fit(self, fitPairs=None, quiet=True):
"""
Fit efficiency curve to values given by 'fitPairs' which should be a list
of energy<->efficiency pairs. (See hdtv.util.Pairs())
'energies' and 'efficiencies' may be a list of ufloats
"""
# TODO: Unify this with the energy calibration fitter
if fitPairs is not None:
#self.fitInput = fitPairs
self._fitInput = fitPairs
E = array.array("d")
delta_E = array.array("d")
eff = array.array("d")
delta_eff = array.array("d")
EN = array.array("d")
effN = array.array("d")
# map(energies.append(self.fitInput[0]), self.fitInput)
# map(efficiencies.append(self.fitInput[1]), self.fitInput)
hasXerrors = False
# Convert energies to array needed by ROOT
try:
list(map(lambda x: E.append(x[0].nominal_value), self._fitInput))
list(map(lambda x: delta_E.append(x[0].std_dev), self._fitInput))
list(map(lambda x: EN.append(0.0), self._fitInput))
hasXerrors = True
except AttributeError: # energies does not seem to be ufloat list
list(map(lambda x: E.append(x[0]), self._fitInput))
list(map(lambda x: delta_E.append(0.0), self._fitInput))
# Convert efficiencies to array needed by ROOT
try:
list(map(lambda x: eff.append(x[1].nominal_value), self._fitInput))
list(map(lambda x: delta_eff.append(x[1].std_dev), self._fitInput))
list(map(lambda x: effN.append(0.0), self._fitInput))
except AttributeError: # energies does not seem to be ufloat list
list(map(lambda x: eff.append(x[1]), self._fitInput))
list(map(lambda x: delta_eff.append(0.0), self._fitInput))
# if fit has errors we first fit without errors to get good initial values
# if hasXerrors == True:
# print "Fit parameter without errors included:"
#self.TGraphWithoutErrors = TGraphErrors(len(E), E, eff, EN, effN)
#fitWithoutErrors = self.TGraphWithoutErrors.Fit(self.id, "SF")
hdtv.ui.msg("Fit parameter with errors included:")
# Preliminary normalization
# if self._doNorm:
# self.norm = 1 / max(efficiencies)
# for i in range(len(eff)):
# eff[i] *= self.norm
# delta_eff[i] *= self.norm
#self.TF1.SetRange(0, max(E) * 1.1)
# self.TF1.SetParameter(0, 1) # Unset normalization for fitting
self.TGraph = TGraphErrors(len(E), E, eff, delta_E, delta_eff)
# Do the fit
fitopts = "0" # Do not plot
if hasXerrors:
# We must use the iterative fitter (minuit) to take x errors
# into account.
fitopts += "F"
hdtv.ui.info(
"switching to non-linear fitter (minuit) for x error weighting"
)
if quiet:
fitopts += "Q"
fitopts += "S" # Additional fitinfo returned needed for ROOT5.26 workaround below
fitreturn = self.TGraph.Fit(self.id, fitopts)
try:
# Workaround for checking the fitstatus in ROOT 5.26 (TFitResultPtr
# does not cast properly to int)
fitstatus = fitreturn.Get().Status()
except AttributeError: # This is for ROOT <= 5.24, where fit returns an int
fitstatus = int(fitreturn)
if fitstatus != 0:
# raise RuntimeError, "Fit failed"
hdtv.ui.msg("Fit failed")
# # Final normalization
# if self._doNorm:
# self.normalize()
# Get parameter
for i in range(self._numPars):
self.parameter[i] = self.TF1.GetParameter(i)
# Get covariance matrix
tvf = TVirtualFitter.GetFitter()
## cov = tvf.GetCovarianceMatrix()
for i in range(0, self._numPars):
for j in range(0, self._numPars):
self.fCov[i][j] = tvf.GetCovarianceMatrixElement(i, j)
## self.fCov[i][j] = cov[i * self._numPars + j]
return self.parameter
def doFit(FunctionType,hMassNEW,g,fFitXmin,fFitXmax,fNbins,xbins,fLabel):
print ""
print ""
print ""
print ""
if( FunctionType==-2 ):
nPar=2
BKGfit = TF1("BKGfit%i"%FunctionType," [0] / ( TMath::Power(x/13000,[1]) )",fFitXmin,fFitXmax)
# signal region:
if fLabel.find("WW") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0,1.98680e-09)
BKGfit.SetParameter(1,8.66744e+00)
elif fLabel.find("WW") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,1.13062e-07)
BKGfit.SetParameter(1,7.85359e+00)
elif fLabel.find("WZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0,7.09589e-09)
BKGfit.SetParameter(1,8.54541e+00)
elif fLabel.find("WZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,1.50873e-07)
BKGfit.SetParameter(1,8.03595e+00)
elif fLabel.find("ZZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0,1.82729e-09)
BKGfit.SetParameter(1,8.92196e+00)
elif fLabel.find("ZZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,6.82046e-08)
BKGfit.SetParameter(1,8.08889e+00)
# sideband region:
if fLabel.find("WW") != -1 and fLabel.find("HP") != -1 and fLabel.find("SB")!=-1:
BKGfit.SetParameter(0,1.98680e-09)
BKGfit.SetParameter(1,8.66744e+00)
elif fLabel.find("W") != -1 and fLabel.find("LP") != -1 and fLabel.find("SB")!=-1:
BKGfit.SetParameter(0,1.13062e-07)
BKGfit.SetParameter(1,7.85359e+00)
elif fLabel.find("Z") != -1 and fLabel.find("HP") != -1 and fLabel.find("SB")!=-1:
BKGfit.SetParameter(0,7.09589e-09)
BKGfit.SetParameter(1,8.54541e+00)
elif fLabel.find("Z") != -1 and fLabel.find("LP") != -1 and fLabel.find("SB")!=-1:
BKGfit.SetParameter(0,1.50873e-07)
BKGfit.SetParameter(1,8.03595e+00)
if( FunctionType==0 ):
print "Fitting three parameter default function!"
nPar=3
BKGfit = TF1("BKGfit%i"%FunctionType,"( [0]*TMath::Power(1-x/13000,[1]) ) / ( TMath::Power(x/13000,[2]) )",fFitXmin,fFitXmax)
# signal region:
if fLabel.find("WW") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0,7.68629e-09 )
BKGfit.SetParameter(1,8.35270e+00 )
BKGfit.SetParameter(2, 0.)
#BKGfit.SetParameter(0, 9.01237e-12)
#BKGfit.SetParameter(1,-1.58982e+01)
#BKGfit.SetParameter(2, 1.02766e+01)
BKGfit.SetParLimits(2,7,20)
elif fLabel.find("WW") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,3.75544e-07)
BKGfit.SetParameter(1,3.37623e+00)
BKGfit.SetParameter(2,7.49005e+00)
elif fLabel.find("WZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0,7.09589e-09)
BKGfit.SetParameter(1,8.54541e+00)
BKGfit.SetParameter(2, 8.87103e+00)
elif fLabel.find("WZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,1.50874e-07)
BKGfit.SetParameter(1,8.03595e+00)
BKGfit.SetParameter(2,7.37063e+00)
elif fLabel.find("ZZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0,1.06871e-08)
BKGfit.SetParameter(1,1.03718e+00)
BKGfit.SetParameter(2,8.38073e+00)
BKGfit.SetParLimits(2, 8.2e+00 ,8.4e+00 )
BKGfit.SetParLimits(1, 1.03718e+00 , 1.03718e+00)
elif fLabel.find("ZZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,6.82046e-08)
BKGfit.SetParameter(2,8.08889e+00)
BKGfit.SetParameter(1,0.00000000 )
# sideband region:
if fLabel.find("WW") != -1 and fLabel.find("HP") != -1 and fLabel.find("SB")!=-1:
BKGfit.SetParameter(0, 9.01237e-12)
BKGfit.SetParameter(1,-1.58982e+01)
BKGfit.SetParameter(2, 1.02766e+01)
elif fLabel.find("WW") != -1 and fLabel.find("LP") != -1 and fLabel.find("SB")!=-1:
BKGfit.SetParameter(0,3.75544e-07)
BKGfit.SetParameter(1,3.37623e+00)
BKGfit.SetParameter(2,7.49005e+00)
elif fLabel.find("ZZ") != -1 and fLabel.find("HP") != -1 and fLabel.find("SB")!=-1:
BKGfit.SetParameter(0,1.82729e-09)
BKGfit.SetParameter(1,8.92196e+00)
BKGfit.SetParameter(2,9.48734e+00)
elif fLabel.find("ZZ") != -1 and fLabel.find("LP") != -1 and fLabel.find("SB")!=-1:
BKGfit.SetParameter(0,6.82046e-08)
BKGfit.SetParameter(2,8.08889e+00)
BKGfit.SetParameter(1,0.00000000 )
elif( FunctionType==1 ):
print "Fitting four parameter default function!"
nPar=4
BKGfit = TF1("BKGfit%i"%FunctionType,"( [0]*TMath::Power(1-x/13000,[1]) ) / ( TMath::Power(x/13000,[2]+[3]*log(x/13000)) )",fFitXmin,fFitXmax)
# signal region:
if fLabel.find("WW") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 2.45156e-02) #6.48238e-13 )
BKGfit.SetParameter(1, 2.60131e+01 ) #1.11517e+00 )
BKGfit.SetParameter(2, 1.17317e+00 ) #1.02840e+01 )
BKGfit.SetParameter(3, -8.19261e-01 ) #8.45161e+03 )
#BKGfit.FixParameter(3,0 )
#BKGfit.SetParLimits(1, 2.60131e+01, 2.60131e+01)
#BKGfit.SetParLimits(2,0,10)
#BKGfit.SetParLimits(3,-8.19261e-01,-8.19261e-01)
#BKGfit.SetParLimits(0,3.01257e-12,9.01257e-12)
elif fLabel.find("WW") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,3.72696e-07)
BKGfit.SetParameter(1,3.35574e+00)
BKGfit.SetParameter(2,7.49238e+00)
BKGfit.SetParameter(3,-4.10905e-07)
elif fLabel.find("WZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0,7.09589e-09)
BKGfit.SetParameter(1,8.54541e+00)
BKGfit.SetParameter(2, 8.87103e+00)
BKGfit.SetParameter(3,-8.14899e-02)
elif fLabel.find("WZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,1.50873e-07)
BKGfit.SetParameter(1,8.03595e+00)
BKGfit.SetParameter(2,7.96865e+00)
BKGfit.SetParameter(3,9.66574e-02)
elif fLabel.find("ZZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 5.90181e-02)# 1.82729e-09)
BKGfit.SetParameter(1, 2.12431e+01)# 8.92196e+00)
BKGfit.SetParameter(2,-1.00178e+00)# 4.76973e+00)
BKGfit.SetParameter(3,-1.54769e+00)#-7.21037e-01)
#BKGfit.SetParLimits(3, -1.54769e+00,-1.54769e+00)
#BKGfit.SetParLimits(2,-1.00178e+00,-1.00178e+00 )
#BKGfit.SetParLimits(1,2.12431e+01,2.12431e+01 )
#BKGfit.SetParLimits(0, 5.90181e-02, 5.90181e-02)
elif fLabel.find("ZZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,6.82046e-08)
BKGfit.SetParameter(2,8.08889e+00)
BKGfit.SetParameter(1,0.00000000 )
BKGfit.SetParameter(3,0.00000000 )
# sideband region:
if fLabel.find("WW") != -1 and fLabel.find("HP") != -1 and fLabel.find("SB")!=-1:
BKGfit.SetParameter(0, 9.01257e-12)
BKGfit.SetParameter(1,-1.09759e+01)
BKGfit.SetParameter(2, 8.55014e+00)
BKGfit.SetParameter(3,-1 )# -2.54954e-01)
BKGfit.SetParLimits(3,-10,0)
elif fLabel.find("WW") != -1 and fLabel.find("LP") != -1 and fLabel.find("SB")!=-1:
BKGfit.SetParameter(0,3.72696e-07)
BKGfit.SetParameter(1,3.35574e+00)
BKGfit.SetParameter(2,7.49238e+00)
BKGfit.SetParameter(3,-4.10905e-07)
elif fLabel.find("ZZ") != -1 and fLabel.find("HP") != -1 and fLabel.find("SB")!=-1 :
BKGfit.SetParameter(0,1.82729e-09)
BKGfit.SetParameter(1,8.92196e+00)
BKGfit.SetParameter(2, 4.76973e+00)
BKGfit.SetParameter(3,-7.21037e-01)
elif fLabel.find("ZZ") != -1 and fLabel.find("LP") != -1 and fLabel.find("SB")!=-1:
BKGfit.SetParameter(0,6.82046e-08)
BKGfit.SetParameter(2,8.08889e+00)
BKGfit.SetParameter(1,0.00000000 )
BKGfit.SetParameter(3,0.00000000 )
elif( FunctionType==3 ):
print "Fitting four parameter alternate function!"
nPar=4
BKGfit = TF1("BKGfit%i"%FunctionType,"( [0]*TMath::Power(1-x/13000 + [3]*TMath::Power(x/13000,2),[1]) ) / ( TMath::Power(x/13000,[2]) )",fFitXmin,fFitXmax)
# signal region:
if fLabel.find("WW") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0,2.21926e-13)
BKGfit.SetParameter(1,2.25629e+00)
BKGfit.SetParameter(2,1.15338e+01)
BKGfit.SetParameter(3,2.11319e+02)
# BKGfit.SetParLimits(0,0.21926e-13,4.21926e-13) #WWHP
elif fLabel.find("WW") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,1.44775e-05 )
BKGfit.SetParameter(1,2.05225e+01 )
BKGfit.SetParameter(2,6.50590e+00 )
BKGfit.SetParameter(3,1.86664e+00 )
elif fLabel.find("WZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0,1.19345e-05)
BKGfit.SetParameter(1,3.73975e+01)
BKGfit.SetParameter(2,6.58579e+00)
BKGfit.SetParameter(3,2.48983e+00)
elif fLabel.find("WZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,1.36722e-06)
BKGfit.SetParameter(1,6.63254e+00)
BKGfit.SetParameter(2,7.37687e+00)
BKGfit.SetParameter(3,3.28215e-01)
elif fLabel.find("ZZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 2.45629e-09 )
BKGfit.SetParameter(1,-4.58690e+00 )
BKGfit.SetParameter(2, 8.80612e+00 )
BKGfit.SetParameter(3, 2.05830e+00 )
#BKGfit.SetParLimits(2, 8.80612e+00, 8.80612e+00)
#BKGfit.SetParLimits(3, 2.05830e+00, 2.05830e+00)
elif fLabel.find("ZZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 1.74162e-06)
BKGfit.SetParameter(1,-8.47400e-01)
BKGfit.SetParameter(2, 7.05095e+00)
BKGfit.SetParameter(3, 1.79446e+02)
#sideband region:
if fLabel.find("W") != -1 and fLabel.find("HP") != -1 and fLabel.find("SB")!=-1:
BKGfit.SetParameter(0,2.21926e-13)
BKGfit.SetParameter(1,2.25629e+00)
BKGfit.SetParameter(2,1.15338e+01)
BKGfit.SetParameter(3,2.11319e+02)
elif fLabel.find("W") != -1 and fLabel.find("LP") != -1 and fLabel.find("SB")!=-1:
BKGfit.SetParameter(0,1.44775e-05 )
BKGfit.SetParameter(1,2.05225e+01 )
BKGfit.SetParameter(2,6.50590e+00 )
BKGfit.SetParameter(3,1.86664e+00 )
elif fLabel.find("Z") != -1 and fLabel.find("HP") != -1 and fLabel.find("SB")!=-1:
BKGfit.SetParameter(0, 1.82729e-09)
BKGfit.SetParameter(1, 8.92196e+00)
BKGfit.SetParameter(2, 4.76973e+00)
BKGfit.SetParameter(3,-7.21037e-01)
elif fLabel.find("Z") != -1 and fLabel.find("LP") != -1 and fLabel.find("SB")!=-1:
BKGfit.SetParameter(0, 1.74162e-06)
BKGfit.SetParameter(1,-8.47400e-01)
BKGfit.SetParameter(2, 7.05095e+00)
BKGfit.SetParameter(3, 1.79446e+02)
stopProgram=1;
for loop in range (0,10):
r = hMassNEW.Fit("BKGfit%i"%FunctionType,"BISR","",fFitXmin,fFitXmax)
fitStatus = int(r)
print "fit status : %d" % fitStatus
if(fitStatus==0):
stopProgram=0
# r.Print("V")
break
c1 = TCanvas("c3","fit",600,400)
hMassNEW.Draw();
suffix ="WHP";
if fLabel.find('W')!=-1:
suffix ="W"
if fLabel.find('Z')!=-1:
suffix ="Z"
if fLabel.find('Z')!=-1:
suffix ="Z"
if fLabel.find('WW')!=-1:
suffix ="WW"
if fLabel.find('WZ')!=-1:
suffix ="WZ"
if fLabel.find('ZZ')!=-1:
suffix ="WZ"
if fLabel.find('HP')!=-1:
suffix +="HP"
if fLabel.find('LP')!=-1:
suffix +="LP"
c1.SaveAs("ftest_2016/test_"+suffix+".pdf");
#if(stopProgram==1):
# print "######################"
# print"######################"
# print "ERROR : Fit %i failed!!!!" %FunctionType
# print "######################"
# print "######################"
# sys.exit()
#Create a histogram to hold the confidence intervals
# histoCI=TH1D("histoCI","", fNbins,xbins)
# # histoCI=TH1D()
# histoCI.SetTitle("Fitted histogram with .95 conf. band")
# (TVirtualFitter.GetFitter()).GetConfidenceIntervals(histoCI)
# #Now the "hint" histogram has the fitted function values as the
# #bin contents and the confidence intervals as bin errors
# histoCI.SetStats(kFALSE)
# histoCI.SetFillColor(kRed+2)
# histoCI.SetLineColor(kRed+2)
# histoCI.SetFillStyle(3354)
histoCI = TGraphErrors(fNbins)
histoCI.SetTitle("Fitted line with .95 conf. band")
for i in range (0, fNbins):
histoCI.SetPoint(i, g.GetX()[i], 0)
#Compute the confidence intervals at the x points of the created graph
(TVirtualFitter.GetFitter()).GetConfidenceIntervals(histoCI)
# //Now the "grint" graph contains function values as its y-coordinates
# //and confidence intervals as the errors on these coordinates
# //Draw the graph, the function and the confidence intervals
histoCI.SetLineColor(0)
histoCI.SetLineWidth(-802)
histoCI.SetFillStyle(3002)
histoCI.SetFillColor(2)
NumberOfVarBins = 0
NumberOfObservations_VarBin = 0
chi2_VarBin = 0.
chi2_VarBin_notNorm = 0.
hist_fit_residual_vsMass = TH1D("hist_fit_residual_vsMass","hist_fit_residual_vsMass",fNbins,xbins)
for bin in range (1,hMassNEW.GetNbinsX()):
if( hMassNEW.GetXaxis().GetBinCenter(bin)>=fFitXmin and hMassNEW.GetXaxis().GetBinCenter(bin)<=fFitXmax ):
NumberOfVarBins += 1
data = hMassNEW.GetBinContent(bin)
# data = g.Integral(hMassNEW.GetXaxis().GetBinLowEdge(bin) , hMassNEW.GetXaxis().GetBinUpEdge(bin) )
err_data_low = g.GetErrorYlow(bin-1)
err_data_high= g.GetErrorYhigh(bin-1)
fit = BKGfit.Integral(hMassNEW.GetXaxis().GetBinCenter(bin) , hMassNEW.GetXaxis().GetBinCenter(bin) )
fit = fit / ( hMassNEW.GetBinWidth(bin) )
# fit = BKGfit.Eval(hMassNEW.GetBinCenter(bin)) #yields same results
if(fit > data):
err_tot = err_data_high
else:
err_tot = err_data_low
if err_tot!=0:
fit_residual = (data - fit) / err_tot
else:
print "Warning: err_tot = 0 !"
fit_residual = 0
err_fit_residual = 1
if (hMassNEW.GetBinContent(bin)>0):
NumberOfObservations_VarBin+=1
print "(data-fit)**2 = "+str(pow( (data - fit) , 2 ))+" err_tot : "+str(err_tot)
print chi2_VarBin
chi2_VarBin += pow( (data - fit) , 2 ) / pow( err_tot , 2 )
chi2_VarBin_notNorm += pow( (data - fit) , 2 )
hist_fit_residual_vsMass.SetBinContent(bin,fit_residual)
hist_fit_residual_vsMass.SetBinError(bin,err_fit_residual)
ndf_VarBin = NumberOfObservations_VarBin - nPar -1 #ndof
return [chi2_VarBin_notNorm,ndf_VarBin,chi2_VarBin,BKGfit,hist_fit_residual_vsMass,nPar,histoCI]
gRho00VsPromptFrac[iPt].SetPoint(iPoint, promptFrac, rho00)
gRho00VsPromptFrac[iPt].SetPointError(iPoint, uncPromptFrac,
uncPromptFrac, uncRho00,
uncRho00)
gRho00VsNonPromptFrac[iPt].SetPoint(iPoint, nonpromptFrac, rho00)
gRho00VsNonPromptFrac[iPt].SetPointError(iPoint, uncNonpromptFrac,
uncNonpromptFrac, uncRho00,
uncRho00)
cRho00VsPromptFrac.cd(iPt + 1).DrawFrame(
0., 0., 1., 0.6, ';#it{f}_{prompt};#it{#rho}_{00}')
lineNoPol.Draw('same')
gRho00VsPromptFrac[iPt].Fit(f'fRho00VsPromptFrac_pt{ptMin}-{ptMax}', '0')
TVirtualFitter.GetFitter().GetConfidenceIntervals(ciRho00VsPromptFrac[iPt],
0.68)
ciRho00VsPromptFrac[iPt].DrawCopy('same')
fRho00VsPromptFrac[iPt].Draw('same')
gRho00VsPromptFrac[iPt].Draw('PZ')
lat.DrawLatex(0.5, 0.4, f'{ptMin} < #it{{p}}_{{T}} < {ptMax} GeV/#it{{c}}')
lat.DrawLatex(
0.5, 0.3,
f'#chi^{{2}}/ndf = {fRho00VsPromptFrac[iPt].GetChisquare():0.2f}/{fRho00VsPromptFrac[iPt].GetNDF()}'
)
cRho00VsNonPromptFrac.cd(iPt + 1).DrawFrame(
0., 0., 1., 0.6, ';#it{f}_{non-prompt};#it{#rho}_{00}')
lineNoPol.Draw('same')
gRho00VsNonPromptFrac[iPt].Fit(f'fRho00VsNonPromptFrac_pt{ptMin}-{ptMax}',
'0')
TVirtualFitter.GetFitter().GetConfidenceIntervals(
def fit(self):
"""
Some comments:
-> The object 'back_fit' is fitted using only the side-bands
-> The object 'tot_fit' is fitted to the entire extracted inv-mass distribution
-> The object 'back_refit' is not used for fitting but only stores the background fit
parameters from fitted 'tot_fit'
-> The object 'sig_fit' is not used for fitting but only stores the signal fit
parameters from fitted 'tot_fit'
"""
self.logger.info("Do fitting")
TVirtualFitter.SetDefaultFitter("Minuit")
# Prepare what can be prepared
npar_sig = self.sig_fit_func.GetNpar()
# Need to do it that way since background functions have a few dummy parameters
npar_bkg = self.tot_fit_func.GetNpar() - npar_sig
self.logger.info("Initial parameters for signal fit are")
print(f"mean = {self.mean}\nsigma = {self.sigma}")
self.logger.debug("fit background (just side bands)")
self.histo_to_fit.Fit(self.bkg_sideband_fit_func, ("R,%s,+,0" % (self.fit_options)))
# Prepare a function to store the signal parameters which will finally be extracted
# from the total fit. So this is just a helper for now
minForSig = self.mean - self.nsigma_sideband * self.sigma
maxForSig = self.mean + self.nsigma_sideband * self.sigma
binForMinSig = self.histo_to_fit.FindBin(minForSig)
binForMaxSig = self.histo_to_fit.FindBin(maxForSig)
sum_tot = 0.
sumback = 0.
for ibin in range(binForMinSig, binForMaxSig + 1):
sum_tot += self.histo_to_fit.GetBinContent(ibin)
sumback += self.bkg_sideband_fit_func.Eval(self.histo_to_fit.GetBinCenter(ibin))
integsig = Double((sum_tot - sumback) * (self.histo_to_fit.GetBinWidth(1)))
self.sig_fit_func.SetParameter(0, integsig)
self.sig_fit_func.SetParameter(1, self.mean)
self.sig_fit_func.SetParameter(2, self.sigma)
self.logger.info("fit all (signal + background)")
self.tot_fit_func.SetLineColor(4)
parmin = Double()
parmax = Double()
for ipar in range(0, npar_sig):
self.tot_fit_func.SetParameter(ipar + npar_bkg, self.sig_fit_func.GetParameter(ipar))
self.sig_fit_func.GetParLimits(ipar, parmin, parmax)
self.tot_fit_func.SetParLimits(ipar + npar_bkg, parmin, parmax)
for ipar in range(0, npar_bkg):
self.tot_fit_func.SetParameter(ipar, self.bkg_sideband_fit_func.GetParameter(ipar))
self.bkg_fit_func.SetParameter(ipar, self.bkg_sideband_fit_func.GetParameter(ipar))
if self.fix_mean:
# Mass peak would be fixed to what user sets
self.tot_fit_func.FixParameter(npar_bkg + 1, self.mean)
if self.fix_sigma is True:
# Sigma would be fixed to what the fit to MC gives
self.tot_fit_func.FixParameter(npar_bkg + 2,
self.tot_fit_func.GetParameter(npar_bkg + 2))
self.histo_to_fit.Fit(self.tot_fit_func, ("R,%s,+,0" % (self.fit_options)))
for ipar in range(0, npar_bkg):
self.bkg_tot_fit_func.SetParameter(ipar, self.tot_fit_func.GetParameter(ipar))
self.bkg_tot_fit_func.SetParError(ipar, self.tot_fit_func.GetParameter(ipar))
for ipar in range(npar_bkg, (npar_bkg + npar_sig)):
self.sig_fit_func.SetParameter(ipar - npar_bkg, self.tot_fit_func.GetParameter(ipar))
self.sig_fit_func.SetParError(ipar - npar_bkg, self.tot_fit_func.GetParError(ipar))
self.derive_yields()
# Check the signal fit
error = self.update_check_signal_fit()
if error != "":
self.logger.error("Signal fit probably bad for following reasons:\n%s", error)
self.fitted = True
self.fit_success = (error == "")
return self.fit_success
def doFit(FunctionType, hMassNEW, g, fFitXmin, fFitXmax, fNbins, xbins,
fLabel):
print ""
print ""
print ""
print ""
if (FunctionType == -2):
nPar = 2
BKGfit = TF1("BKGfit%i" % FunctionType,
" [0] / ( TMath::Power(x/13000,[1]) )", fFitXmin,
fFitXmax)
if fLabel.find("qW") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 8.10795e-06)
BKGfit.SetParameter(1, 7.30736e+00)
elif fLabel.find("qW") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 1.13062e-07)
BKGfit.SetParameter(1, 7.85359e+00)
elif fLabel.find("qZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 7.09589e-09)
BKGfit.SetParameter(1, 8.54541e+00)
elif fLabel.find("qZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 1.50873e-07)
BKGfit.SetParameter(1, 8.03595e+00)
elif fLabel.find("qV") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 7.09589e-09)
BKGfit.SetParameter(1, 8.54541e+00)
elif fLabel.find("qV") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 1.50873e-07)
BKGfit.SetParameter(1, 8.03595e+00)
if (FunctionType == 0):
print "Fitting three parameter default function!"
nPar = 3
BKGfit = TF1(
"BKGfit%i" % FunctionType,
"( [0]*TMath::Power(1-x/13000,[1]) ) / ( TMath::Power(x/13000,[2]) )",
fFitXmin, fFitXmax)
if fLabel.find("qW") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 1.21219e-04)
BKGfit.SetParameter(1, 6.16942e+00)
BKGfit.SetParameter(2, 6.66576e+00)
elif fLabel.find("qW") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 1.47860e-03)
BKGfit.SetParameter(1, 7.17033e+00)
BKGfit.SetParameter(2, 6.00169e+00)
elif fLabel.find("qZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 5.79282e-05)
BKGfit.SetParameter(1, 4.53461e+00)
BKGfit.SetParameter(2, 7.03951e+00)
elif fLabel.find("qZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 9.62968e-04)
BKGfit.SetParameter(1, 6.99994e+00)
BKGfit.SetParameter(2, 6.09152e+00)
elif fLabel.find("qV") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 9.62968e-04)
BKGfit.SetParameter(1, 6.99994e+00)
BKGfit.SetParameter(2, 6.09152e+00)
elif fLabel.find("qV") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 9.62968e-04)
BKGfit.SetParameter(1, 6.99994e+00)
BKGfit.SetParameter(2, 6.09152e+00)
elif (FunctionType == 1):
print "Fitting four parameter default function!"
nPar = 4
BKGfit = TF1(
"BKGfit%i" % FunctionType,
"( [0]*TMath::Power(1-x/13000,[1]) ) / ( TMath::Power(x/13000,[2]+[3]*log(x/13000)) )",
fFitXmin, fFitXmax)
if fLabel.find("qW") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 1.14515e-04)
BKGfit.SetParameter(1, 5.88071e+00)
BKGfit.SetParameter(2, 6.65185e+00)
BKGfit.SetParameter(3, -1.09416e-02)
elif fLabel.find("qW") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 2.17738e-03)
BKGfit.SetParameter(1, 6.84330e+00)
BKGfit.SetParameter(2, 6.43714e+00)
BKGfit.SetParameter(3, 6.55647e-02)
elif fLabel.find("qZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 5.79282e-05)
BKGfit.SetParameter(1, 4.53461e+00)
BKGfit.SetParameter(2, 7.03951e+00)
BKGfit.SetParameter(3, 2.28311e-02)
elif fLabel.find("qZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 8.67191e-03)
BKGfit.SetParameter(1, 8.18650e+00)
BKGfit.SetParameter(2, 5.25282e+00)
BKGfit.SetParameter(3, -1.79962e-01)
elif fLabel.find("qV") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 3.47742e-05)
BKGfit.SetParameter(1, 4.11131e+00)
BKGfit.SetParameter(2, 6.97258e+00)
BKGfit.SetParameter(3, 0.0000)
elif fLabel.find("qV") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 9.62968e-04)
BKGfit.SetParameter(1, 6.99994e+00)
BKGfit.SetParameter(2, 6.09152e+00)
BKGfit.SetParameter(3, 0.0000)
elif (FunctionType == 3):
print "Fitting five parameter default function!"
nPar = 5
BKGfit = TF1(
"BKGfit%i" % FunctionType,
"( [0]*TMath::Power(1-x/13000,[1]) ) / ( TMath::Power(x/13000,[2]+[3]*log(x/13000)+[4]*TMath::Power(log(x/13000),2)) )",
fFitXmin, fFitXmax)
if fLabel.find("qW") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 1.06734e-04)
BKGfit.SetParameter(1, 7.17665e+00)
BKGfit.SetParameter(2, 7.68365e+00)
BKGfit.SetParameter(3, 7.09362e-01)
BKGfit.SetParameter(4, 1.35015e-01)
elif fLabel.find("qW") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 1.20333e-04)
BKGfit.SetParameter(1, 7.69352e+00)
BKGfit.SetParameter(2, 7.79226e+00)
BKGfit.SetParameter(3, 8.15310e-01)
BKGfit.SetParameter(4, 1.55179e-01)
elif fLabel.find("qZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 4.36219e-05)
BKGfit.SetParameter(1, 3.19544e+00)
BKGfit.SetParameter(2, 7.08912e+00)
BKGfit.SetParameter(3, -7.33087e-01)
BKGfit.SetParameter(4, -1.01751e-02)
BKGfit.SetParLimits(4, -1.01751e-02, -1.01751e-02)
elif fLabel.find("qZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 9.62968e-04)
BKGfit.SetParameter(1, 6.99994e+00)
BKGfit.SetParameter(2, 6.09152e+00)
elif fLabel.find("qV") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 3.47742e-05)
BKGfit.SetParameter(1, 4.11131e+00)
BKGfit.SetParameter(2, 6.97258e+00)
BKGfit.SetParameter(3, 0.0000)
BKGfit.SetParameter(4, 0.0000)
elif fLabel.find("qV") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 9.62968e-04)
BKGfit.SetParameter(1, 6.99994e+00)
BKGfit.SetParameter(2, 6.09152e+00)
BKGfit.SetParameter(3, 0.0000)
BKGfit.SetParameter(4, 0.0000)
elif (FunctionType == 4):
print "Fitting four parameter alternate function!"
nPar = 4
BKGfit = TF1(
"BKGfit%i" % FunctionType,
"( [0]*TMath::Power(1-x/13000 + [3]*TMath::Power(x/13000,2),[1]) ) / ( TMath::Power(x/13000,[2]) )",
fFitXmin, fFitXmax)
if fLabel.find("qW") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 8.10795e-06)
BKGfit.SetParameter(1, 4.67052e+00)
BKGfit.SetParameter(2, 7.30736e+00)
BKGfit.SetParameter(3, 8.09331e+00)
elif fLabel.find("qW") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 5.14674e-03)
BKGfit.SetParameter(1, 8.54997e+00)
BKGfit.SetParameter(2, 5.98285e+00)
BKGfit.SetParameter(3, 1.89857e-01)
elif fLabel.find("qZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 3.64825e-05)
BKGfit.SetParameter(1, 2.78348e+00)
BKGfit.SetParameter(2, 7.17321e+00)
BKGfit.SetParameter(3, -9.83737e-01)
BKGfit.SetParLimits(0, -0.1, 1)
BKGfit.SetParLimits(1, 0, 5)
BKGfit.SetParLimits(2, 0, 10)
BKGfit.SetParLimits(3, -1, 1)
elif fLabel.find("qZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 1.06334e-03)
BKGfit.SetParameter(1, 2.96443e+00)
BKGfit.SetParameter(2, 6.36638e+00)
BKGfit.SetParameter(3, -1.72927e+00)
elif fLabel.find("qV") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 3.47742e-05)
BKGfit.SetParameter(1, 4.11131e+00)
BKGfit.SetParameter(2, 6.97258e+00)
BKGfit.SetParameter(3, 0.0000)
elif fLabel.find("qV") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 2.10360e+01) # 3.61006e+01)
BKGfit.SetParameter(1, -3.67723e+00) #-3.14135e+00)
BKGfit.SetParameter(2, 2.66588e+00) # 2.31349e+00 )
BKGfit.SetParameter(3, 1.31511e+02) # 1.32204e+02)
#BKGfit.SetParameter(3,0.0000)
elif (FunctionType == 5):
print "Fitting five parameter alternate function!"
nPar = 5
BKGfit = TF1(
"BKGfit%i" % FunctionType,
"( [0]*TMath::Power(1-x/13000 + [3]*TMath::Power(x/13000,2),[1]) ) / ( TMath::Power(x/13000,[2]+[4]*log(x/13000)) )",
fFitXmin, fFitXmax)
if fLabel.find("qW") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 4.99818e-07)
BKGfit.SetParameter(1, -8.15352e+00)
BKGfit.SetParameter(2, 9.17744e+00)
BKGfit.SetParameter(3, 2.49453e+00)
BKGfit.SetParameter(4, 2.98004e-01)
elif fLabel.find("qW") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 1.47860e-03)
BKGfit.SetParameter(1, 7.17031e+00)
BKGfit.SetParameter(2, 6.00169e+00)
BKGfit.SetParameter(3, 5.03605e+00)
BKGfit.SetParameter(4, -4.77303e-01)
elif fLabel.find("qZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 4.36216e-05)
BKGfit.SetParameter(1, 3.19548e+00)
BKGfit.SetParameter(2, 7.08912e+00)
BKGfit.SetParameter(3, -7.32868e-01)
BKGfit.SetParameter(4, -1.01744e-02)
BKGfit.SetParLimits(0, -0.1, 1)
BKGfit.SetParLimits(1, 0, 5)
BKGfit.SetParLimits(2, 0, 10)
BKGfit.SetParLimits(3, -1, 1)
BKGfit.SetParLimits(4, -1, 1)
elif fLabel.find("qZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 2.93921e-03)
BKGfit.SetParameter(1, 6.35371e+00)
BKGfit.SetParameter(2, 5.86362e+00)
BKGfit.SetParameter(3, -1.98677e-01)
BKGfit.SetParameter(4, -8.50249e-02)
elif fLabel.find("qV") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 3.47742e-05) #3.47742e-05)
BKGfit.SetParameter(1, 4.11131e+00) #4.11131e+00)
BKGfit.SetParameter(2, 6.97258e+00) #6.97258e+00)
BKGfit.SetParameter(3, 0.0) #0.0000)
BKGfit.SetParameter(4, 0.0) #0.0000)
#BKGfit.SetParLimits(4,-5,5)
#BKGfit.SetParLimits(3,-1.794,10)
elif fLabel.find("qV") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(
0, 1.70901e-04
) # 8.96766e-04) 1 p0 1.70900e-04
BKGfit.SetParameter(
1, 3.00147e+01
) # 6.70628e+00) 2 p1 3.00147e+01
BKGfit.SetParameter(
2, 1.06146e+01
) # 6.11149e+00) 3 p2 1.06146e+01
BKGfit.SetParameter(
3, 1.62392e+00
) #1.23384e+00 )# -7.82019e-02) 4 p3 1.62392e+00
BKGfit.SetParameter(
4, 1.29753e+00
) # 0.0000) 5 p4 1.29753e+00
elif (FunctionType == 3):
print "Fitting four parameter alternate function!"
nPar = 4
BKGfit = TF1(
"BKGfit%i" % FunctionType,
"( [0]*TMath::Power(1-x/13000 + [3]*TMath::Power(x/13000,2),[1]) ) / ( TMath::Power(x/13000,[2]) )",
fFitXmin, fFitXmax)
stopProgram = 1
for loop in range(0, 10):
r = hMassNEW.Fit("BKGfit%i" % FunctionType, "ISRBL", "", fFitXmin,
fFitXmax)
fitStatus = int(r)
print "Inetgral of first bin:"
print BKGfit.Integral(fFitXmin, 1118)
print hMassNEW.GetBinContent(hMassNEW.GetXaxis().FindBin(1110))
print "fit status : %d" % fitStatus
if (fitStatus == 0):
stopProgram = 0
# r.Print("V")
#break
#if(stopProgram==1):
#print "######################"
#print"######################"
#print "ERROR : Fit %i failed!!!!" %FunctionType
#print "######################"
#print "######################"
#sys.exit()
#Create a histogram to hold the confidence intervals
# histoCI=TH1D("histoCI","", fNbins,xbins)
# # histoCI=TH1D()
# histoCI.SetTitle("Fitted histogram with .95 conf. band")
# (TVirtualFitter.GetFitter()).GetConfidenceIntervals(histoCI)
# #Now the "hint" histogram has the fitted function values as the
# #bin contents and the confidence intervals as bin errors
# histoCI.SetStats(kFALSE)
# histoCI.SetFillColor(kRed+2)
# histoCI.SetLineColor(kRed+2)
# histoCI.SetFillStyle(3354)
histoCI = TGraphErrors(fNbins)
histoCI.SetTitle("Fitted line with .95 conf. band")
for i in range(0, fNbins):
histoCI.SetPoint(i, g.GetX()[i], 0)
#Compute the confidence intervals at the x points of the created graph
(TVirtualFitter.GetFitter()).GetConfidenceIntervals(histoCI)
# //Now the "grint" graph contains function values as its y-coordinates
# //and confidence intervals as the errors on these coordinates
# //Draw the graph, the function and the confidence intervals
histoCI.SetLineColor(0)
histoCI.SetLineWidth(-802)
histoCI.SetFillStyle(3002)
histoCI.SetFillColor(2)
NumberOfVarBins = 0
NumberOfObservations_VarBin = 0
chi2_VarBin = 0.
chi2_VarBin_notNorm = 0.
hist_fit_residual_vsMass = TH1D("hist_fit_residual_vsMass",
"hist_fit_residual_vsMass", fNbins, xbins)
for bin in range(1, hMassNEW.GetNbinsX()):
if (hMassNEW.GetXaxis().GetBinLowEdge(bin) >= fFitXmin
and hMassNEW.GetXaxis().GetBinUpEdge(bin) <= fFitXmax):
#print "computing chi2"
#print "bin: " + str(bin) + " content " + str(hMassNEW.GetBinContent(bin))
NumberOfVarBins += 1
data = hMassNEW.GetBinContent(bin)
# data = g.Integral(hMassNEW.GetXaxis().GetBinLowEdge(bin) , hMassNEW.GetXaxis().GetBinUpEdge(bin) )
err_data_low = g.GetErrorYlow(bin - 1)
err_data_high = g.GetErrorYhigh(bin - 1)
fit = BKGfit.Integral(hMassNEW.GetXaxis().GetBinLowEdge(bin),
hMassNEW.GetXaxis().GetBinUpEdge(bin))
fit = fit / (hMassNEW.GetBinWidth(bin))
# fit = BKGfit.Eval(hMassNEW.GetBinCenter(bin)) #yields same results
if (fit > data):
err_tot = err_data_high
else:
err_tot = err_data_low
if err_tot != 0:
fit_residual = (data - fit) / err_tot
else:
fit_residual = 0
err_fit_residual = 1
if (hMassNEW.GetBinContent(bin) > 0):
NumberOfObservations_VarBin += 1
chi2_VarBin += pow((data - fit), 2) / pow(err_tot, 2)
chi2_VarBin_notNorm += pow((data - fit), 2)
hist_fit_residual_vsMass.SetBinContent(bin, fit_residual)
hist_fit_residual_vsMass.SetBinError(bin, err_fit_residual)
ndf_VarBin = NumberOfObservations_VarBin - nPar - 1 #ndof
return [
chi2_VarBin_notNorm, ndf_VarBin, chi2_VarBin, BKGfit,
hist_fit_residual_vsMass, nPar, histoCI
]
gMassPValue.append(TGraph(0))
gMassNsigma.append(TGraph(0))
SetObjectStyle(gMassPValue[iPt], color=kRed + 1)
SetObjectStyle(gMassNsigma[iPt], color=kRed + 1)
hMassToFit = hMassSel[iPt].Clone()
for iBin in range(hMassToFit.GetNbinsX()): #remove signal region
if mParticle - 0.02 < hMassToFit.GetBinCenter(iBin +
1) < mParticle + 0.02:
hMassToFit.SetBinContent(iBin + 1, 0)
hMassToFit.Fit(fMass[iPt], 'RE0')
hConfIntBkg.append(
TH1F(f'hConfIntMassKKMCPt{ptMin}_{ptMax}', '',
hMassSel[iPt].GetNbinsX(), minMass, maxMass))
TVirtualFitter.GetFitter().GetConfidenceIntervals(hConfIntBkg[iPt], 0.683)
SetObjectStyle(hConfIntBkg[iPt],
color=kBlack,
fillalpha=0.3,
markerstyle=0)
if iPt == 0:
legFit.AddEntry(hMassSel[iPt], 'ML selected', 'lp')
legFit.AddEntry(hConfIntBkg[iPt], f'{args.fitfunc} fit', 'fl')
legFit.AddEntry(line, f'{particleTitle} mass', 'l')
cMassFit.append(
TCanvas(f'cMassFitPt{ptMin*10:.0f}_{ptMax*10:.0f}', '', 1200, 400))
cMassFit[iPt].Divide(3, 1)
cMassFit[iPt].cd(1).DrawFrame(minMass, 0., maxMass,
hMassSel[iPt].GetMaximum() * 1.5,
f'{ptMin:.0f} < #it{{p}}_{{T}} < {ptMax:.0f} GeV/#it{{c}}')
hMassKKBkgSubData[iPt].GetYaxis().SetTitle(f'Normalised Counts-{subLabel}')
# fit distribution with voigt function
fMassKKMC.append(TF1(f'fMassKKMCPt{iPt}', VoigtFunc, 0.98, 1.08, 4))
fMassKKMC[iPt].SetLineColor(kRed + 1)
fMassKKMC[iPt].SetParameters(hMassKKBkgSubMC[iPt].GetBinWidth(1), massPhi,
0.0015, widthPhi)
fMassKKMC[iPt].FixParameter(0, hMassKKBkgSubMC[iPt].GetBinWidth(1))
fMassKKMC[iPt].SetParLimits(1, massPhi * 0.5, massPhi * 2)
fMassKKMC[iPt].SetParLimits(2, 0., 0.1)
fMassKKMC[iPt].FixParameter(3, widthPhi)
resMC = hMassKKBkgSubMC[iPt].Fit(fMassKKMC[iPt], 'ER0SQ')
hConfIntMassKKMC.append(
TH1F(f'hConfIntMassKKMC_{iPt}', '', 500, 0.98, 1.08))
TVirtualFitter.GetFitter().GetConfidenceIntervals(hConfIntMassKKMC[iPt],
0.68)
SetObjectStyle(hConfIntMassKKMC[iPt],
color=kRed + 1,
alpha=0.3,
markerstyle=0)
fMassKKData.append(TF1(f'fMassKKDataPt{iPt}', VoigtFunc, 0.98, 1.08, 4))
fMassKKData[iPt].SetLineColor(kAzure + 4)
fMassKKData[iPt].SetParameters(hMassKKBkgSubData[iPt].GetBinWidth(1),
massPhi, 0.0015, widthPhi)
fMassKKData[iPt].FixParameter(0, hMassKKBkgSubData[iPt].GetBinWidth(1))
fMassKKData[iPt].SetParLimits(1, massPhi * 0.5, massPhi * 2)
fMassKKData[iPt].SetParLimits(2, 0., 0.1)
fMassKKData[iPt].FixParameter(3, widthPhi)
resData = hMassKKBkgSubData[iPt].Fit(fMassKKData[iPt], 'ER0SQ')
hConfIntMassKKData.append(
e1 = round(e1,4)
# value of the first parameter
p2 = fit_x.GetParameter(1)
p2 = round(p2,4)
# error of the first parameter
e2 = fit_x.GetParError(1)
e2 = round(e2,4)
print "Chi2 = " + str(chi2)
print "P1 = " + str(p1)
print "E1 = " + str(e1)
TVirtualFitter.GetFitter().GetConfidenceIntervals(grint)
grint.SetLineColor(1)
grint.SetFillColorAlpha(16,0.65)
grint.GetHistogram().SetTitle(" ");
grint.GetHistogram().GetXaxis().SetTitle("Top P_{T} (GeV)")
if(THEORY == "PWHGP8" ):
grint.GetHistogram().GetYaxis().SetTitle("DATA/Powheg+Pythia8")
elif(THEORY == "Kidonakis"):
grint.GetHistogram().GetYaxis().SetTitle("DATA/aN^3LO")
elif(THEORY == "Mitov"):
grint.GetHistogram().GetYaxis().SetTitle("DATA/NNLO")
elif(THEORY == "Lipka"):
def doFit(FunctionType,hMassNEW,hMass,g,fFitXmin,fFitXmax,fNbins,xbins):
if( FunctionType==-2 ):
nPar=2
BKGfit = TF1("BKGfit%i"%FunctionType," [0] / ( TMath::Power(x/13000,[1]) )",fFitXmin,fFitXmax)
# BKGfit.SetParameter(0,8.63367e-11)
# BKGfit.SetParameter(1,9.39072e+00)
if( FunctionType==0 ):
print "Fitting three parameter default function!"
nPar=3
BKGfit = TF1("BKGfit%i"%FunctionType,"( [0]*TMath::Power(1-x/13000,[1]) ) / ( TMath::Power(x/13000,[2]) )",fFitXmin,fFitXmax)
# SIGfit = TF1("SIGfit%i"%FunctionType,"(TMath::Gaus(x,[3]))",fFitXmin,fFitXmax)
# BKGfit.SetParameter(0,8.63367e-11)
# BKGfit.SetParameter(1,9.39072e+00)
# BKGfit.SetParameter(0,1.03079e-08)
# BKGfit.SetParameter(1,7.86584e+00)
# BKGfit.SetParameter(2,1.42945e+01)
elif( FunctionType==1 ):
print "Fitting four parameter default function!"
nPar=4
BKGfit = TF1("BKGfit%i"%FunctionType,"( [0]*TMath::Power(1-x/13000,[1]) ) / ( TMath::Power(x/13000,[2]+[3]*log(x/13000)) )",fFitXmin,fFitXmax)
BKGfit.SetParameter(0,8.63367e-11)
BKGfit.SetParameter(1,9.39072e+00)
BKGfit.SetParameter(0,4.07887e-18)
BKGfit.SetParLimits(0,4.15001e-18-4.07887e-18,4.15001e-18+4.07887e-18)
BKGfit.SetParameter(1,7.86584e+00)
BKGfit.SetParameter(2,1.42945e+01)
BKGfit.SetParameter(3,-5.1)
elif( FunctionType==2 ):
print "Fitting five parameter function!"
nPar=5
# BKGfit = TF1("BKGfit%i"%FunctionType,"( [0]*TMath::Power(1-x/13000,[1]) ) / ( TMath::Power(x/13000,[2]+[3]*log(x/13000)+[4]*TMath::Power(log(x/13000),2)) )",fFitXmin,fFitXmax)
BKGfit = TF1("BKGfit%i"%FunctionType,"( [0]*TMath::Power(1-x/13000,[1])*(1+[4]*x/13000) ) / ( TMath::Power(x/13000,[2]+[3]*log(x/13000)) )",fFitXmin,fFitXmax)
BKGfit.SetParameter(0,4.07887e-18)
BKGfit.SetParLimits(0,4.15001e-18-4.07887e-18,4.15001e-18+4.07887e-18)
BKGfit.SetParameter(1,7.86584e+00)
BKGfit.SetParameter(2,1.42945e+01)
BKGfit.SetParameter(3,-5.1000699000)
elif( FunctionType==3 ):
print "Fitting four parameter alternate function!"
nPar=4
BKGfit = TF1("BKGfit%i"%FunctionType,"( [0]*TMath::Power(1-x/13000 + [3]*TMath::Power(x/13000,2),[1]) ) / ( TMath::Power(x/13000,[2]) )",fFitXmin,fFitXmax)
# BKGfit.SetParameter(0,3.61696e-06) #VVHP
# BKGfit.SetParameter(1,-1.29729e+00 )
# BKGfit.SetParameter(2,5.93603e+00 )
# BKGfit.SetParameter(3,3.02948e+02 )
stopProgram=1;
for loop in range (0,10):
r = hMassNEW.Fit("BKGfit%i"%FunctionType,"ISR","",fFitXmin,fFitXmax)
fitStatus = int(r)
print "fit status : %d" % fitStatus
if(fitStatus==0):
stopProgram=0
# r.Print("V")
break
if(stopProgram==1):
print "######################"
print"######################"
print "ERROR : Fit %i failed!!!!" %FunctionType
print "######################"
print "######################"
sys.exit()
NumberOfVarBins = 0
NumberOfObservations_VarBin = 0
chi2_VarBin = 0.
chi2_VarBin_ALL = 0.
chi2_VarBin_notNorm = 0.
chi2_VarBin_notNorm_ALL = 0.
chi2_VarBin_notNorm_over5 = 0.
chi2_VarBin_over5 = 0.
NumberOfObservations_VarBin_over5 = 0.
#Create a histogram to hold the confidence intervals
# histoCI=TH1D("histoCI","", fNbins,xbins)
# # histoCI=TH1D()
# histoCI.SetTitle("Fitted histogram with .95 conf. band")
# (TVirtualFitter.GetFitter()).GetConfidenceIntervals(histoCI)
# #Now the "hint" histogram has the fitted function values as the
# #bin contents and the confidence intervals as bin errors
# histoCI.SetStats(kFALSE)
# histoCI.SetFillColor(kRed+2)
# histoCI.SetLineColor(kRed+2)
# histoCI.SetFillStyle(3354)
histoCI = TGraphErrors(fNbins)
histoCI.SetTitle("Fitted line with .95 conf. band")
for i in range (0, fNbins):
histoCI.SetPoint(i, g.GetX()[i], 0)
#Compute the confidence intervals at the x points of the created graph
(TVirtualFitter.GetFitter()).GetConfidenceIntervals(histoCI)
# //Now the "grint" graph contains function values as its y-coordinates
# //and confidence intervals as the errors on these coordinates
# //Draw the graph, the function and the confidence intervals
histoCI.SetLineColor(0)
histoCI.SetLineWidth(-802)
histoCI.SetFillStyle(3002)
histoCI.SetFillColor(2)
hist_fit_residual_vsMass = TH1D("hist_fit_residual_vsMass","hist_fit_residual_vsMass",fNbins,xbins)
for bin in range (1,hMassNEW.GetNbinsX()):
if( hMassNEW.GetXaxis().GetBinLowEdge(bin)>=fFitXmin and hMassNEW.GetXaxis().GetBinUpEdge(bin)<=fFitXmax ):
NumberOfVarBins += 1
data = hMassNEW.GetBinContent(bin)
err_data_low = g.GetErrorYlow(bin-1)
err_data_high= g.GetErrorYhigh(bin-1)
fit = BKGfit.Integral(hMassNEW.GetXaxis().GetBinLowEdge(bin) , hMassNEW.GetXaxis().GetBinUpEdge(bin) )
fit = fit / ( hMassNEW.GetBinWidth(bin) )
# fit = BKGfit.Eval(hMassNEW.GetBinCenter(bin)) #yields same results
if(fit > data):
err_tot = err_data_high
else:
err_tot = err_data_low
# if(err_tot==0):continue
fit_residual = (data - fit) / err_tot
err_fit_residual = 1
chi2_VarBin_notNorm_ALL += pow( (data - fit) , 2 ) #For alternate method from Yongjie, not skipping emty bins
chi2_VarBin_ALL += pow( (data - fit) , 2 ) / pow( err_tot , 2 )
if (hMassNEW.GetBinContent(bin)>0.0000000000001):
NumberOfObservations_VarBin+=1
chi2_VarBin += pow( (data - fit) , 2 ) / pow( err_tot , 2 )
chi2_VarBin_notNorm += pow( (data - fit) , 2 ) #For dijet method, skipping emty bins
print "Mass = %f , Bin = %i , data = %f, fit = %f, rss = %f , +=rss= %f" %(hMassNEW.GetBinCenter(bin),bin,data,fit,(pow( (data - fit) , 2 )),chi2_VarBin_notNorm)
if (hMassNEW.GetBinContent(bin)*hMassNEW.GetBinWidth(bin)>=5):
NumberOfObservations_VarBin_over5+=1
chi2_VarBin_over5 += pow( (data - fit) , 2 ) / pow( err_tot , 2 )
chi2_VarBin_notNorm_over5+= pow( (data - fit) , 2 ) #For F-test, skipping bins with less than 5 entries
hist_fit_residual_vsMass.SetBinContent(bin,fit_residual)
hist_fit_residual_vsMass.SetBinError(bin,err_fit_residual)
ndf_VarBin = NumberOfObservations_VarBin - nPar -1 #ndof for >0 bins
ndf_VarBin_over5 = NumberOfObservations_VarBin_over5 - nPar -1 #ndof for >5 bins
return [chi2_VarBin_notNorm,ndf_VarBin,chi2_VarBin,BKGfit,hist_fit_residual_vsMass,nPar,chi2_VarBin_ALL,chi2_VarBin_notNorm_ALL,histoCI,ndf_VarBin_over5,chi2_VarBin_notNorm_over5]
def doFit(FunctionType,hMassNEW,g,fFitXmin,fFitXmax,fNbins,xbins,fLabel):
print ""
print ""
if( FunctionType==-2 ):
nPar=2
BKGfit = TF1("BKGfit%i"%FunctionType," [0] / ( TMath::Power(x/13000,[1]) )",fFitXmin,fFitXmax)
BKGfit.SetParameter(0, 1.76820e-08 )#4.92405e-09 ) #WWHP
BKGfit.SetParameter(1, 8.25223e+00 )#7.90584e+00 ) #WWHP
if fLabel.find("qV") != -1:
BKGfit.SetParameter(0, 3.72282e-12 )
BKGfit.SetParameter(1, -5.30842e+00 )
if fLabel.find("WW") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0,8.22440e-09) #WWLP
BKGfit.SetParameter(1,8.48110e+00)
if fLabel.find("WW") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,3.54934e-07 ) #WWLP
if( FunctionType==0 ):
print "Fitting three parameter default function!"
nPar=3
BKGfit = TF1("BKGfit%i"%FunctionType,"( [0]*TMath::Power(1-x/13000,[1]) ) / ( TMath::Power(x/13000,[2]) )",fFitXmin,fFitXmax)
if fLabel.find("qV") != -1:
BKGfit.SetParameter(0, 3.72282e-12 )
BKGfit.SetParameter(1, -5.30842e+00 )
BKGfit.SetParameter(2, 1.82771e+01 )
if fLabel.find("qW")!=-1 or fLabel.find("qZ")!=-1:
BKGfit.SetParameter(0,5.83605e-09 ) #WWHP
BKGfit.SetParameter(1,4.83899e-01 ) #WWHP
BKGfit.SetParameter(2,7.85446e+00 ) #WWHP
if fLabel.find("WW") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0,3.09869e-05)
BKGfit.SetParameter(1,2.11957e+01)
BKGfit.SetParameter(2,10)
BKGfit.SetParLimits(2,10.2,20)
if fLabel.find("WW") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 1.91544e-07 ) #WWLP
BKGfit.SetParameter(1,-7.48572e-01 ) #WWHP
BKGfit.SetParameter(2, 8.07181e+00 ) #WWHP
if fLabel.find("ZZ") != -1 and fLabel.find("HP") != -1:
print "ZZHP three paramter default function"
BKGfit.SetParameter(0, 1.78211e-08 ) #ZZHP
BKGfit.SetParameter(1, 8.31381e+00 ) #WWHP
BKGfit.SetParameter(2, 0. ) #3.36325e+00 )
#BKGfit.SetParLimits(0, 1.78211e-08, 1.78211e-08)
#BKGfit.SetParLimits(1, 8.31381e+00 ,8.31381e+00)
#BKGfit.SetParLimits(2, -0.000001, 0.000001 )
if fLabel.find("ZZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,5.83605e-09 ) #WWHP
BKGfit.SetParameter(1,4.83899e-01 ) #WWHP
BKGfit.SetParameter(2,7.85446e+00 ) #WWHP
if fLabel.find("WZ") != -1 and fLabel.find("HP") != -1:
print "WZHP"
BKGfit.SetParameter(0, 1.35257e-08)
BKGfit.SetParameter(1, 8.74474e+00)
#BKGfit.SetParLimits(0,0,1e-07)
#BKGfit.SetParLimits(1,5,8)
BKGfit.SetParLimits(2,8.8,12)
#BKGfit.SetParameter(0,0)#1.76820e-08)#5.83605e-09 ) #WWHP
#BKGfit.SetParameter(1,0)#8.25223e+00)#4.83899e-01 ) #WWHP
#BKGfit.SetParameter(2,0)#7.85446e+00 ) #WWHP
if fLabel.find("WZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,5.83605e-09 ) #WWHP
BKGfit.SetParameter(1,4.83899e-01 ) #WWHP
BKGfit.SetParameter(2,7.85446e+00 ) #WWHP
elif( FunctionType==1 ):
print "Fitting four parameter default function!"
nPar=4
BKGfit = TF1("BKGfit%i"%FunctionType,"( [0]*TMath::Power(1-x/13000,[1]) ) / ( TMath::Power(x/13000,[2]+[3]*log(x/13000)) )",fFitXmin,fFitXmax)
if fLabel.find("qV") != -1:
BKGfit.SetParameter(0, 3.72282e-12 )
BKGfit.SetParameter(1, -5.30842e+00 )
BKGfit.SetParameter(2, 1.82771e+01 )
BKGfit.SetParameter(3, 2.27782e+00 )
if fLabel.find("WZ") != -1 and fLabel.find("HP") != -1:
#BKGfit.SetParLimits(0, 1.57838e+07,1.57838e+08 )
BKGfit.SetParameter(0, 2.93331e-14 )
BKGfit.SetParameter(1, -1.50024e+01 )
BKGfit.SetParameter(2, 1.69784e+01 )
BKGfit.SetParameter(3, 1.41238e+00 )
BKGfit.SetParLimits(3,0,5)
if fLabel.find("WW") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 5.08045e-10 )
BKGfit.SetParameter(1, 8.10729e+00 )
BKGfit.SetParameter(2, 1.28397e+01 )
BKGfit.SetParameter(3, 1.20048e+00 )
BKGfit.SetParLimits(3,-10,0)
BKGfit.SetParLimits(2,10,20)
if fLabel.find("WW") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 1.34232e-02 )#WWLP
BKGfit.SetParameter(1, 1.35564e+01 ) #WWLP
BKGfit.SetParameter(2, 1.29328e+00 )#WWLP
BKGfit.SetParameter(3,-1.12300e+00 )#WWLP
if fLabel.find("WZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 1.40401e-06)#1.45686e-06)#WWLP
BKGfit.SetParameter(1, 2.78968e+00)#2.86174e+00) #WWLP
BKGfit.SetParameter(2, 7.70104e+00)#7.68425e+00) #WWLP
BKGfit.SetParameter(3, -1) #WWLP
BKGfit.SetParLimits(3,-10,-0.005)
if fLabel.find("ZZ") != -1 and fLabel.find("HP") != -1:
# BKGfit.SetParLimits(0, 9.66361e-08,9.66361e-06 )
BKGfit.SetParameter(0, 9.11215e-09)
BKGfit.SetParameter(1, 2.02554e+01)
BKGfit.SetParameter(2, 1.36639e+01)
BKGfit.SetParameter(3, 0.) #1.77601e+00)
#BKGfit.SetParLimits(3,0.,0.)
if fLabel.find("ZZ") != -1 and fLabel.find("LP") != -1:
#BKGfit.SetParLimits(0, 1)#9.66361e-08,9.66361e-04 )
BKGfit.SetParameter(0, 6.27404e-07 )#6.92267e-04 )
BKGfit.SetParameter(1, 2.69667e+00 )#1.18029e+01 )
BKGfit.SetParameter(2, 7.73807e+00 )# 3.50726e+00)
BKGfit.SetParameter(3, 1 )#-6.96555e-01 )
BKGfit.SetParLimits(3,0,2)
elif( FunctionType==2 ):
print "Fitting five parameter default function!"
nPar=5
BKGfit = TF1("BKGfit%i"%FunctionType,"( [0]*TMath::Power(1-x/13000,[1]) ) / ( TMath::Power(x/13000,[2]+[3]*log(x/13000)+[4]*TMath::Power(log(x/13000),2)) )",fFitXmin,fFitXmax)
if fLabel.find("qW") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 5.30325e-05)
# BKGfit.SetParameter(1, 4.87226e+00)
# BKGfit.SetParameter(2, 6.71810e+00)
# BKGfit.SetParameter(3,0.001)
elif fLabel.find("qW") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 1.69726e-03)
BKGfit.SetParameter(1, 7.40452e+00)
BKGfit.SetParameter(2, 5.93207e+00)
BKGfit.SetParameter(3,-7.75527e-03)
elif fLabel.find("qZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0,3.47742e-05)
BKGfit.SetParameter(1,4.11131e+00)
BKGfit.SetParameter(2,6.97258e+00)
BKGfit.SetParameter(3,0.0000)
BKGfit.SetParameter(4,0.0000)
elif fLabel.find("qZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,1.06977e-01 )
BKGfit.SetParameter(1, 9.45383e+00)
BKGfit.SetParameter(2, 3.83836e+00)
BKGfit.SetParameter(3,-5.80838e-01)
BKGfit.SetParameter(4,-3.49361e-02)
elif fLabel.find("qV") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0,3.47742e-05)
BKGfit.SetParameter(1,4.11131e+00)
BKGfit.SetParameter(2,6.97258e+00)
BKGfit.SetParameter(3,0.0000)
BKGfit.SetParameter(4,0.0000)
elif fLabel.find("qV") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,9.62968e-04)
BKGfit.SetParameter(1,6.99994e+00)
BKGfit.SetParameter(2,6.09152e+00)
BKGfit.SetParameter(3,0.0000)
BKGfit.SetParameter(4,0.0000)
elif( FunctionType==3 ):
print "Fitting four parameter alternate function!"
nPar=4
BKGfit = TF1("BKGfit%i"%FunctionType,"( [0]*TMath::Power(1-x/13000 + [3]*TMath::Power(x/13000,2),[1]) ) / ( TMath::Power(x/13000,[2]) )",fFitXmin,fFitXmax)
if fLabel.find("q") != -1:
BKGfit.SetParameter(0, 3.45939e+02 )
BKGfit.SetParameter(1,-4.45967e+00 )
BKGfit.SetParameter(2, 1.31413e+00 )
BKGfit.SetParameter(3, 1.91782e+02 )
if fLabel.find("qV")!=-1 and fLabel.find("HP")!=-1:
BKGfit.SetParameter(0, 3.45939e+02 )
BKGfit.SetParameter(1,-4.45967e+00 )
BKGfit.SetParameter(2, 1.31413e+00 )
BKGfit.SetParameter(3, 0.0)
if fLabel.find("WW") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 7.16992e-05 )
BKGfit.SetParameter(1, -3.00473e+00 )
#BKGfit.SetParameter(2, 5.35092e+00 )
#BKGfit.SetParameter(3, 9.42630e+01 )
BKGfit.SetParLimits(2,5,10)
BKGfit.SetParLimits(3,-15,1)
if fLabel.find("WW") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 4.48531e-07 )# 1.47497e+08) #WWLP
BKGfit.SetParameter(1, -7.96655e-01)#5.01221e+01 ) #WWLP
BKGfit.SetParameter(2, 7.43952e+00 )#-1.21037e+01) #WWLP
BKGfit.SetParameter(3,0.)#-3.10164e+00) #WWLP
if fLabel.find("ZZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 4.48531e-07 ) #ZZLP
BKGfit.SetParameter(1, -7.96655e-01 ) #WWLP
BKGfit.SetParameter(2, 7.43952e+00 ) #WWLP
BKGfit.SetParameter(3,0. ) #ZZLP
if fLabel.find("ZZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 9.66361e-07) #ZZHP
BKGfit.SetParameter(1, -1.15120e+01)
BKGfit.SetParameter(2, 6.63440e+00)
BKGfit.SetParameter(3, 1.04241e+01)
if fLabel.find("WZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 1.31274e-02)
BKGfit.SetParameter(1, -2.42314e+00 )
BKGfit.SetParameter(2, 4.70224e+00 )
BKGfit.SetParameter(3, 5.41163e+02 )
if fLabel.find("WZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,4.48531e-07 ) #WWLP
BKGfit.SetParameter(1,-7.96655e-01) #WWLP
BKGfit.SetParameter(2,7.43952e+00 ) #WWLP
BKGfit.SetParameter(3, 0.0) #WWLP
elif( FunctionType==4 ):
print "Fitting five parameter alternate function!"
nPar=5
BKGfit = TF1("BKGfit%i"%FunctionType,"( [0]*TMath::Power(1-x/13000 + [3]*TMath::Power(x/13000,2),[1]) ) / ( TMath::Power(x/13000,[2]+[4]*log(x/13000)) )",fFitXmin,fFitXmax)
if fLabel.find("qW") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0, 7.93174e-05)
BKGfit.SetParameter(1,-9.05071e+00)
BKGfit.SetParameter(2, 5.02176e+00)
BKGfit.SetParameter(3, 5.03605e+00)
BKGfit.SetParameter(4,-4.77303e-01)
BKGfit.SetParameter(5, 7.93174e-05)
elif fLabel.find("qW") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0,1.47860e-03)
BKGfit.SetParameter(1,7.17031e+00)
BKGfit.SetParameter(2,6.00169e+00)
elif fLabel.find("qZ") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0,3.47742e-05)
BKGfit.SetParameter(1,4.11131e+00)
BKGfit.SetParameter(2,6.97258e+00)
BKGfit.SetParameter(3,0.0000)
BKGfit.SetParameter(4,0.0000)
elif fLabel.find("qZ") != -1 and fLabel.find("LP") != -1:
BKGfit.SetParameter(0, 3.08932e-04)
BKGfit.SetParameter(1, -4.02134e+00)
BKGfit.SetParameter(2, 7.00824e+00)
BKGfit.SetParameter(3, 5.03328e+00)
BKGfit.SetParameter(4,-4.77303e-01)
elif fLabel.find("qV") != -1 and fLabel.find("HP") != -1:
BKGfit.SetParameter(0,3.47742e-05)
BKGfit.SetParameter(1,4.11131e+00)
BKGfit.SetParameter(2,6.97258e+00)
BKGfit.SetParameter(3,0.0000)
BKGfit.SetParameter(4,0.0000)
elif fLabel.find("qV") != -1 and fLabel.find("LP") != -1:
#BKGfit.SetParameter(0, 8.96766e-04)
#BKGfit.SetParameter(1, 6.70628e+00)
#BKGfit.SetParameter(2, 6.11149e+00)
BKGfit.SetParameter(0,3.47742e-05)
BKGfit.SetParameter(1,4.11131e+00)
BKGfit.SetParameter(2,6.97258e+00)
BKGfit.SetParameter(3, 0.0)
BKGfit.SetParameter(4,0.0)
stopProgram=1;
for loop in range (0,30):
r = hMassNEW.Fit("BKGfit%i"%FunctionType,"ISRB","",fFitXmin,fFitXmax)
# r = hMassNEW.Fit("BKGfit%i"%FunctionType,"LFISR","",fFitXmin,fFitXmax)
fitStatus = int(r)
print "fit status : %d" % fitStatus
#if(fitStatus==0 and loop > 3):
#stopProgram=0
## r.Print("V")
#break
#if(stopProgram==1):
#print "######################"
#print"######################"
#print "ERROR : Fit %i failed!!!!" %FunctionType
#print "######################"
#print "######################"
#sys.exit()
#Create a histogram to hold the confidence intervals
# histoCI=TH1D("histoCI","", fNbins,xbins)
# # histoCI=TH1D()
# histoCI.SetTitle("Fitted histogram with .95 conf. band")
# (TVirtualFitter.GetFitter()).GetConfidenceIntervals(histoCI)
# #Now the "hint" histogram has the fitted function values as the
# #bin contents and the confidence intervals as bin errors
# histoCI.SetStats(kFALSE)
# histoCI.SetFillColor(kRed+2)
# histoCI.SetLineColor(kRed+2)
# histoCI.SetFillStyle(3354)
histoCI = TGraphErrors(fNbins)
histoCI.SetTitle("Fitted line with .95 conf. band")
for i in range (0, fNbins):
histoCI.SetPoint(i, g.GetX()[i], 0)
#Compute the confidence intervals at the x points of the created graph
(TVirtualFitter.GetFitter()).GetConfidenceIntervals(histoCI)
# //Now the "grint" graph contains function values as its y-coordinates
# //and confidence intervals as the errors on these coordinates
# //Draw the graph, the function and the confidence intervals
histoCI.SetLineColor(0)
histoCI.SetLineWidth(-802)
histoCI.SetFillStyle(3002)
histoCI.SetFillColor(2)
NumberOfVarBins = 0
NumberOfObservations_VarBin = 0
chi2_VarBin = 0.
chi2_VarBin_notNorm = 0.
hist_fit_residual_vsMass = TH1D("hist_fit_residual_vsMass","hist_fit_residual_vsMass",fNbins,xbins)
for bin in range (1,hMassNEW.GetNbinsX()):
if( hMassNEW.GetXaxis().GetBinCenter(bin)>=fFitXmin and hMassNEW.GetXaxis().GetBinCenter(bin)<=fFitXmax ):
NumberOfVarBins += 1
data = hMassNEW.GetBinContent(bin)
# data = g.Integral(hMassNEW.GetXaxis().GetBinLowEdge(bin) , hMassNEW.GetXaxis().GetBinUpEdge(bin) )
err_data_low = g.GetErrorYlow(bin-1)
err_data_high= g.GetErrorYhigh(bin-1)
fit = BKGfit.Integral(hMassNEW.GetXaxis().GetBinCenter(bin) , hMassNEW.GetXaxis().GetBinCenter(bin) )
fit = fit / ( hMassNEW.GetBinWidth(bin) )
# fit = BKGfit.Eval(hMassNEW.GetBinCenter(bin)) #yields same results
if(fit > data):
err_tot = err_data_high
else:
err_tot = err_data_low
fit_residual = (data - fit) / err_tot
err_fit_residual = 1
if (hMassNEW.GetBinContent(bin)>0):
NumberOfObservations_VarBin+=1
chi2_VarBin += pow( (data - fit) , 2 ) / pow( err_tot , 2 )
chi2_VarBin_notNorm += pow( (data - fit) , 2 )
hist_fit_residual_vsMass.SetBinContent(bin,fit_residual)
hist_fit_residual_vsMass.SetBinError(bin,err_fit_residual)
ndf_VarBin = NumberOfObservations_VarBin - nPar -1 #ndof
return [chi2_VarBin_notNorm,ndf_VarBin,chi2_VarBin,BKGfit,hist_fit_residual_vsMass,nPar,histoCI]
ay = array('f', (5.0935, 2.1777, 0.2089, -2.3949, -2.4457, -3.0430, -2.2731, -2.0706, -1.6231, -2.5605, -0.7703, -0.3055, 1.6817, 1.8728, 3.6586, 3.2353, 4.2520, 5.2550, 3.8766, 4.2890 ) )
ey=np.array(len(data_x)*[0.5],dtype=np.float)
ex=np.array(len(data_y)*[0],dtype=np.float)
# ... and pass to TGraphErros object
gr=TGraphErrors(nPoints,data_x,data_y,ex,ey)
gr.SetTitle("TGraphErrors mit Fit")
gr.Draw("AP");
Pol=["pol1","pol2","pol3","pol4","pol5","pol6","pol7"]
# Polifit
for i in range(len(polynom)):
gr.Fit(polynom[i],"V")
c1.Update()
gr.Draw('AP')
fitrp = TVirtualFitter.GetFitter()
nPar = fitrp.GetNumberTotalParameters()
covmat = TMatrixD(nPar, nPar,fitrp.GetCovarianceMatrix())
print('The Covariance Matrix is: ')
covmat.Print()
cormat = TMatrixD(covmat)
for i in range(nPar):
for j in range(nPar):
cormat[i][j] = cormat[i][j] / (np.sqrt(covmat[i][i]) * np.sqrt(covmat[j][j]))
print('The Correlation Matrix is: ')
cormat.Print()
raw_input('Press <ret> to continue -> ')
#Legendre Poly
for i in range(len(legendre)):
gr.Fit(legendre[i],"V")
def doFit(FunctionType, hMassNEW, g, fFitXmin, fFitXmax, fNbins, xbins):
print ""
print ""
if (FunctionType == -2):
nPar = 2
BKGfit = TF1("BKGfit%i" % FunctionType,
" [0] / ( TMath::Power(x/13000,[1]) )", fFitXmin,
fFitXmax)
# BKGfit.SetParameter(0,43557e-16 )
if (FunctionType == 0):
print "Fitting three parameter default function!"
nPar = 3
BKGfit = TF1(
"BKGfit%i" % FunctionType,
"( [0]*TMath::Power(1-x/13000,[1]) ) / ( TMath::Power(x/13000,[2]) )",
fFitXmin, fFitXmax)
# BKGfit.SetParameter(0,3.43557e-16 )
# BKGfit.SetParameter(1,-4.20539e+01)
# BKGfit.SetParameter(2,1.29180e+01)
elif (FunctionType == 1):
print "Fitting four parameter default function!"
nPar = 4
BKGfit = TF1(
"BKGfit%i" % FunctionType,
"( [0]*TMath::Power(1-x/13000,[1]) ) / ( TMath::Power(x/13000,[2]+[3]*log(x/13000)) )",
fFitXmin, fFitXmax)
# BKGfit.SetParameter(0,2.89006e-05 )
# BKGfit.SetParameter(1,1.44811e+01 )
# BKGfit.SetParameter(2,6.08355)
elif (FunctionType == 2):
print "Fitting five parameter function!"
nPar = 5
# BKGfit = TF1("BKGfit%i"%FunctionType,"( [0]*TMath::Power(1-x/13000,[1]) ) / ( TMath::Power(x/13000,[2]+[3]*log(x/13000)+[4]*TMath::Power(log(x/13000),2)) )",fFitXmin,fFitXmax)
BKGfit = TF1(
"BKGfit%i" % FunctionType,
"( [0]*TMath::Power(1-x/13000,[1])*(1+[4]*x/13000) ) / ( TMath::Power(x/13000,[2]+[3]*log(x/13000)) )",
fFitXmin, fFitXmax)
# BKGfit.SetParameter(0,2.89006e-05 )
# BKGfit.SetParameter(1,1.44811e+01 )
# BKGfit.SetParameter(2,6.08355)
elif (FunctionType == 3):
print "Fitting four parameter alternate function!"
nPar = 4
BKGfit = TF1(
"BKGfit%i" % FunctionType,
"( [0]*TMath::Power(1-x/13000 + [3]*TMath::Power(x/13000,2),[1]) ) / ( TMath::Power(x/13000,[2]) )",
fFitXmin, fFitXmax)
BKGfit.SetParameter(0, 5.87345e-09) #VVHP
BKGfit.SetParameter(1, 7.86080)
BKGfit.SetParameter(2, 1.29180e+01)
BKGfit.SetParameter(3, .4)
stopProgram = 1
for loop in range(0, 10):
r = hMassNEW.Fit("BKGfit%i" % FunctionType, "ISR", "", fFitXmin,
fFitXmax)
fitStatus = int(r)
print "fit status : %d" % fitStatus
if (fitStatus == 0):
stopProgram = 0
# r.Print("V")
break
if (stopProgram == 1):
print "######################"
print "######################"
print "ERROR : Fit %i failed!!!!" % FunctionType
print "######################"
print "######################"
sys.exit()
#Create a histogram to hold the confidence intervals
# histoCI=TH1D("histoCI","", fNbins,xbins)
# # histoCI=TH1D()
# histoCI.SetTitle("Fitted histogram with .95 conf. band")
# (TVirtualFitter.GetFitter()).GetConfidenceIntervals(histoCI)
# #Now the "hint" histogram has the fitted function values as the
# #bin contents and the confidence intervals as bin errors
# histoCI.SetStats(kFALSE)
# histoCI.SetFillColor(kRed+2)
# histoCI.SetLineColor(kRed+2)
# histoCI.SetFillStyle(3354)
histoCI = TGraphErrors(fNbins)
histoCI.SetTitle("Fitted line with .95 conf. band")
for i in range(0, fNbins):
histoCI.SetPoint(i, g.GetX()[i], 0)
#Compute the confidence intervals at the x points of the created graph
(TVirtualFitter.GetFitter()).GetConfidenceIntervals(histoCI)
# //Now the "grint" graph contains function values as its y-coordinates
# //and confidence intervals as the errors on these coordinates
# //Draw the graph, the function and the confidence intervals
histoCI.SetLineColor(0)
histoCI.SetLineWidth(-802)
histoCI.SetFillStyle(3002)
histoCI.SetFillColor(2)
NumberOfVarBins = 0
NumberOfObservations_VarBin = 0
chi2_VarBin = 0.
chi2_VarBin_notNorm = 0.
hist_fit_residual_vsMass = TH1D("hist_fit_residual_vsMass",
"hist_fit_residual_vsMass", fNbins, xbins)
for bin in range(1, hMassNEW.GetNbinsX()):
if (hMassNEW.GetXaxis().GetBinLowEdge(bin) >= fFitXmin
and hMassNEW.GetXaxis().GetBinUpEdge(bin) <= fFitXmax):
NumberOfVarBins += 1
data = hMassNEW.GetBinContent(bin)
# data = g.Integral(hMassNEW.GetXaxis().GetBinLowEdge(bin) , hMassNEW.GetXaxis().GetBinUpEdge(bin) )
err_data_low = g.GetErrorYlow(bin - 1)
err_data_high = g.GetErrorYhigh(bin - 1)
fit = BKGfit.Integral(hMassNEW.GetXaxis().GetBinLowEdge(bin),
hMassNEW.GetXaxis().GetBinUpEdge(bin))
fit = fit / (hMassNEW.GetBinWidth(bin))
# fit = BKGfit.Eval(hMassNEW.GetBinCenter(bin)) #yields same results
if (fit > data):
err_tot = err_data_high
else:
err_tot = err_data_low
fit_residual = (data - fit) / err_tot
err_fit_residual = 1
if (hMassNEW.GetBinContent(bin) > 0):
NumberOfObservations_VarBin += 1
chi2_VarBin += pow((data - fit), 2) / pow(err_tot, 2)
chi2_VarBin_notNorm += pow((data - fit), 2)
hist_fit_residual_vsMass.SetBinContent(bin, fit_residual)
hist_fit_residual_vsMass.SetBinError(bin, err_fit_residual)
ndf_VarBin = NumberOfObservations_VarBin - nPar - 1 #ndof
return [
chi2_VarBin_notNorm, ndf_VarBin, chi2_VarBin, BKGfit,
hist_fit_residual_vsMass, nPar, histoCI
]
def RunMinuit(sess, nll, feed_dict = None, float_tfpars = None, call_limit = 50000, useGradient = True,
gradient = None, printout = 50, tmpFile = "tmp_result.txt",
runHesse = False, runMinos = False,
options = None, run_metadata = None ) :
"""
Perform MINUIT minimisation of the negative likelihood.
sess : TF session
nll : graph for negitive likelihood to be minimised
feed_dict :
float_tfpars : list of floating parameter to be used in the fit
call_limit : call limit for MINUIT
gradient : external gradient graph. If None and useGradient is not False, will be
calculated internally
useGradient : flag to control the use of analytic gradient while fitting:
None or False : gradient is not used
True or "CHECK" : analytic gradient will be checked with finite elements,
and will be used is they match
"FORCE" : analytic gradient will be used regardless.
printout : Printout frequency
tmpFile : Name of the file with temporary results (updated every time printout is called)
runHesse ; Run HESSE after minimisation
runMinos : Run MINOS after minimisation
options : additional options to pass to TF session run
run_metadata : metadata to pass to TF session run
"""
global cacheable_tensors
if float_tfpars is None :
tfpars = tf.trainable_variables() # Create TF variables
float_tfpars = [ p for p in tfpars if p.floating() ] # List of floating parameters
if useGradient and gradient is None :
gradient = tf.gradients(nll, float_tfpars) # Get analytic gradient
cached_data = {}
fetch_list = []
for i in cacheable_tensors :
if i not in cached_data : fetch_list += [ i ]
if feed_dict :
feeds = dict(feed_dict)
else :
feeds = None
for i in cacheable_tensors :
if i in cached_data : feeds[i] = cached_data[i]
fetch_data = sess.run(fetch_list, feed_dict = feeds ) # Calculate log likelihood
for i,d in zip(fetch_list, fetch_data) :
cached_data[i] = d
if feed_dict :
feeds = dict(feed_dict)
else :
feeds = None
for i in cacheable_tensors :
if i in cached_data : feeds[i] = cached_data[i]
def fcn(npar, gin, f, par, istatus) : # MINUIT fit function
for i in range(len(float_tfpars)) : float_tfpars[i].update(sess, par[i])
f[0] = sess.run(nll, feed_dict = feeds, options = options, run_metadata = run_metadata ) # Calculate log likelihood
if istatus == 2 : # If gradient calculation is needed
dnll = sess.run(gradient, feed_dict = feeds, options = options, run_metadata = run_metadata ) # Calculate analytic gradient
for i in range(len(float_tfpars)) : gin[i] = dnll[i] # Pass gradient to MINUIT
fcn.n += 1
if fcn.n % printout == 0 :
print " Iteration ", fcn.n, ", Flag=", istatus, " NLL=", f[0], ", pars=", sess.run(float_tfpars)
tmp_results = { 'loglh' : f[0], "status" : -1 }
for n,p in enumerate(float_tfpars) :
tmp_results[p.par_name] = (p.prev_value, 0.)
#WriteFitResults(tmp_results, tmpFile)
fcn.n = 0
minuit = TVirtualFitter.Fitter(0, len(float_tfpars)) # Create MINUIT instance
minuit.Clear()
minuit.SetFCN(fcn)
arglist = array.array('d', 10*[0]) # Auxiliary array for MINUIT parameters
for n,p in enumerate(float_tfpars) : # Declare fit parameters in MINUIT
# print "passing parameter %s to Minuit" % p.par_name
step_size = p.step_size
lower_limit = p.lower_limit
upper_limit = p.upper_limit
if not step_size : step_size = 1e-6
if not lower_limit : lower_limit = 0.
if not upper_limit : upper_limit = 0.
minuit.SetParameter(n, p.par_name, p.init_value, step_size, lower_limit, upper_limit)
arglist[0] = 0.5
minuit.ExecuteCommand("SET ERR", arglist, 1) # Set error definition for neg. likelihood fit
if useGradient == True or useGradient == "CHECK" :
minuit.ExecuteCommand("SET GRA", arglist, 0) # Ask analytic gradient
elif useGradient == "FORCE" :
arglist[0] = 1
minuit.ExecuteCommand("SET GRA", arglist, 1) # Ask analytic gradient
arglist[0] = call_limit # Set call limit
minuit.ExecuteCommand("MIGRAD", arglist, 1) # Perform minimisation
minuit.ExecuteCommand("SET NOG", arglist, 0) # Ask no analytic gradient
if runHesse :
minuit.ExecuteCommand("HESSE", arglist, 1)
if runMinos :
minuit.ExecuteCommand("MINOS", arglist, 1)
results = {} # Get fit results and update parameters
for n,p in enumerate(float_tfpars) :
p.update(sess, minuit.GetParameter(n) )
p.fitted_value = minuit.GetParameter(n)
p.error = minuit.GetParError(n)
if runMinos :
eplus = array.array("d", [0.])
eminus = array.array("d", [0.])
eparab = array.array("d", [0.])
globcc = array.array("d", [0.])
minuit.GetErrors(n, eplus, eminus, eparab, globcc)
p.positive_error = eplus[0]
p.negative_error = eminus[0]
results[p.par_name] = ( p.fitted_value, p.error, p.positive_error, p.negative_error )
else :
results[p.par_name] = ( p.fitted_value, p.error)
# Get status of minimisation and NLL at the minimum
maxlh = array.array("d", [0.])
edm = array.array("d", [0.])
errdef = array.array("d", [0.])
nvpar = array.array("i", [0])
nparx = array.array("i", [0])
fitstatus = minuit.GetStats(maxlh, edm, errdef, nvpar, nparx)
# return fit results
results["edm"] = edm[0]
results["loglh"] = maxlh[0]
results["status"] = fitstatus
results["iterations"] = fcn.n
return results
