def get_chi2( regularisation_settings, args, smearing_test=False ):
'''
Takes each tau value, unfolds and refolds, calcs the chi2, the prob of chi2 given ndf (n_bins)
and returns a dictionary of (1-P(Chi2|NDF)) for each tau
For measured test where we only worry about tau=0 outputs tau variables to data frame (+smeared measured values)
'''
h_truth, h_response, h_measured, h_data, h_fakes = regularisation_settings.get_histograms()
# Dont remove any fakes if we are using the true mc distribution
if not args.run_measured_as_data or not args.run_smeared_measured_as_data:
h_data = removeFakes( h_measured, h_fakes, h_data )
variable = regularisation_settings.variable
taus = regularisation_settings.taus_to_test
chi2_ndf = []
for tau in taus:
unfolding = Unfolding(
h_data,
h_truth,
h_measured,
h_response,
fakes = None,#Fakes or no?
method = 'TUnfold',
tau = tau
)
# Cannot refold without first unfolding
h_unfolded_data = unfolding.unfold()
h_refolded_data = unfolding.refold()
# print("Data")
# print (hist_to_value_error_tuplelist(h_data))
# print("Unfolded Data")
# print (hist_to_value_error_tuplelist(h_unfolded_data))
# print("Refolded Data")
# print (hist_to_value_error_tuplelist(h_refolded_data))
regularisation_settings.h_refolded = h_refolded_data
ndf = regularisation_settings.ndf
if args.unfolded_binning:
unfolding.refolded_data = h_refolded_data.rebinned(2)
unfolding.data = h_data.rebinned(2)
ndf = int(regularisation_settings.ndf / 2)
regularisation_settings.h_refolded = unfolding.refolded_data
regularisation_settings.h_data = unfolding.data
if args.create_refold_plots:
plot_data_vs_refold(args, regularisation_settings, tau)
# Calculate the chi2 between refold and unfold
chi2 = unfolding.getUnfoldRefoldChi2()
# Calculate the Prob chi2 given NDF
prob = TMath.Prob( chi2, ndf )
# 1-P(Chi2|NDF)
chi2_ndf.append(1-prob)
# print( tau, chi2, prob, 1-prob )
# Create tau and Chi2 dictionary
d_chi2 = {variable : pd.Series( chi2_ndf )}
d_taus = {'tau' : pd.Series( taus )}
if smearing_test:
d_tau_vars = {
variable : {
'Tau' : tau,
'Chi2' : chi2,
'Prob' : prob,
'1-Prob' : 1-prob,
}
}
df_unfold_tests = tau_vars_to_df(d_tau_vars, regularisation_settings)
return df_unfold_tests
df_chi2 = chi2_to_df(d_chi2, d_taus, regularisation_settings )
return df_chi2
def get_chi2s_of_tau_range( regularisation_settings, args, unfold_test=False ):
'''
Takes each tau value, unfolds and refolds, calcs the chi2, the prob of chi2 given ndf (n_bins)
and returns a dictionary of (1-P(Chi2|NDF)) for each tau
For measured test where we only worry about tau=0 outputs tau variables to data frame (+smeared measured values)
'''
h_truth, h_response, h_measured, h_data, h_fakes = regularisation_settings.get_histograms()
if not args.run_measured_as_data :
h_data = removeFakes( h_measured, h_fakes, h_data )
variable = regularisation_settings.variable
taus = regularisation_settings.taus_to_test
chi2_ndf = []
for tau in taus:
unfolding = Unfolding(
h_data,
h_truth,
h_measured,
h_response,
fakes = None,#Fakes or no?
method = 'TUnfold',
tau = tau
)
# Cannot refold without first unfolding
h_unfolded_data = unfolding.unfold()
h_refolded_data = unfolding.refold()
# print("Data")
# print (hist_to_value_error_tuplelist(h_data))
# print("Unfolded Data")
# print (hist_to_value_error_tuplelist(h_unfolded_data))
# print("Refolded Data")
# print (hist_to_value_error_tuplelist(h_refolded_data))
regularisation_settings.h_refolded = h_refolded_data
ndf = regularisation_settings.ndf
if args.run_refold_plots:
plot_data_vs_refold(args, regularisation_settings, tau)
if args.unfolded_binning:
unfolding.refolded_data = h_refolded_data.rebinned(2)
unfolding.data = h_data.rebinned(2)
ndf = int(regularisation_settings.ndf / 2)
# print("Data")
# print (hist_to_value_error_tuplelist(regularisation_settings.h_data))
# print("Refolded Data")
# print (hist_to_value_error_tuplelist(regularisation_settings.h_refolded))
chi2 = unfolding.getUnfoldRefoldChi2()
prob = TMath.Prob( chi2, ndf )
chi2_ndf.append(1-prob)
# print( tau, chi2, prob, 1-prob )
# Create pandas dictionary
d_chi2 = {variable : pd.Series( chi2_ndf )}
d_taus = {'tau' : pd.Series( taus )}
if unfold_test:
d_tau_vars = {
variable : {
'Tau' : tau,
'Chi2' : chi2,
'Prob' : prob,
'1-Prob' : 1-prob,
}
}
df_unfold_tests = tau_vars_to_df(d_tau_vars, regularisation_settings)
return df_unfold_tests
df_chi2 = chi2_to_df(d_chi2, d_taus, regularisation_settings )
return df_chi2
def main():
config = XSectionConfig(13)
method = 'TUnfold'
# A few different files for testing different inputs
file_for_unfolding = File(config.unfolding_central, 'read')
powheg_herwig_file = File(config.unfolding_powheg_herwig, 'read')
for channel in ['combined', 'muon', 'electron']:
# for variable in config.variables:
for variable in config.variables:
# for variable in ['MET']:
print variable
# tau_value = get_tau_value(config, channel, variable)
# tau_value = 0.000228338590921
tau_value = 0.000
h_truth, h_measured, h_response, h_fakes = get_unfold_histogram_tuple(
inputfile=file_for_unfolding,
variable=variable,
channel=channel,
centre_of_mass=config.centre_of_mass_energy,
ttbar_xsection=config.ttbar_xsection,
luminosity=config.luminosity,
load_fakes=False,
visiblePS=True,
)
# measured = asrootpy(h_response.ProjectionX('px',1))
# print 'Measured from response :',list(measured.y())
# truth = asrootpy(h_response.ProjectionY())
# print 'Truth from response :',list(truth.y())
h_truth_ph, h_measured_ph, h_response_ph, h_fakes_ph = get_unfold_histogram_tuple(
inputfile=powheg_herwig_file,
variable=variable,
channel=channel,
met_type=config.met_type,
centre_of_mass=config.centre_of_mass_energy,
ttbar_xsection=config.ttbar_xsection,
luminosity=config.luminosity,
load_fakes=False,
visiblePS=True,
)
measured = asrootpy(h_response_ph.ProjectionX('px',1))
# print 'Measured from response :',list(measured.y())
measured.SetBinContent(0,0)
truth = asrootpy(h_response_ph.ProjectionY())
# print 'Truth from response :',list(truth.y())
# print 'Truth underflow :',truth.GetBinContent(0),truth.GetBinContent(truth.GetNbinsX()+1)
# Unfold
unfolding = Unfolding( measured,
truth, measured, h_response, None,
method=method, k_value=-1, tau=tau_value)
# unfolded_data = unfolding.closureTest()
# print 'Measured :',list( h_measured.y() )
# h_measured, _ = removeFakes( h_measured, None, h_response)
# for binx in range(0,h_truth.GetNbinsX()+2):
# for biny in range(0,h_truth.GetNbinsX()+2):
# print binx, biny,h_response.GetBinContent(binx,biny)
# print bin,h_truth.GetBinContent(bin)
# print 'Tau :',tau_value
unfolded_results = unfolding.unfold()
# print 'Unfolded :',list( unfolded_results.y() )
# print unfolding.unfoldObject.GetTau()
# print 'Unfolded :',list( unfolded_results.y() )
refolded_results = unfolding.refold()
refolded_results.rebin(2)
measured.rebin(2)
print 'Refolded :',list( refolded_results.y() )
print 'Measured :',list( measured.y() )
# for i in range(1,refolded_results.GetNbinsX()):
# print i,measured.GetBinContent(i),measured.GetBinError(i),abs( measured.GetBinContent(i) - refolded_results.GetBinContent(i) )
pValue = measured.Chi2Test(refolded_results)
print pValue,1-pValue