ROOT Version: 6.10/04 (running on LXPLUS)
Platform: SL6 (LXPLUS)
Compiler: g++ 6.2.0 (presumably, I’m not sure how to get that from root-config)
In PyROOT I’m constructing a ROOT::Fit::Fitter and then attempting to perform a fit with MINOS uncertainties.
Inside a fitting function the fit is constructed and then performed as such
# This is inside of a function that takes certain parameters, which are not defined here. I can add later if wanted.
# Create a fitter
this_Fitter = ROOT.Fit.Fitter()
this_Fitter.Config().MinimizerOptions().SetMinimizerType('Minuit2')
# Minimize is a combination of Migrad and Simplex
# https://root.cern.ch/root/htmldoc/guides/users-guide/FittingHistograms.html#new-rootfit-classes
this_Fitter.Config().MinimizerOptions().SetMinimizerAlgorithm('Minimize')
this_Fitter.Config().MinimizerOptions().SetMaxIterations(1000000)
this_Fitter.Config().MinimizerOptions().SetMaxFunctionCalls(100000)
this_Fitter.Config().MinimizerOptions().SetStrategy(1)
this_Fitter.Config().SetMinosErrors() # Turn MINOS on
this_Fitter.Config().SetUpdateAfterFit() # Set paramters to fit result
# Create function
this_Functor = FitFunction(hist_to_fit, model, fit_range) # This is a class that inherits from ROOT.TPyMultiGenFunction so as to be usable by the fit
aParams = array.array('d', init_params)
aSteps = array.array('d', [0.01] * this_Functor.NDim())
this_Fitter.Config().SetParamsSettings(this_Functor.NDim(), aParams, aSteps)
# some stuff is done here to impose bounds
# Run the fit
this_Fitter.FitFCN(this_Functor)
print('MINOS uncert actually performed? {}'.format(
this_Fitter.CalculateMinosErrors()))
result = ROOT.TFitResult(this_Fitter.Result())
print('fit result paramters')
print([result.Parameter(index) for index in range(result.NPar())])
print('lower, uppper MINOS uncert')
print([(result.LowerError(index), result.UpperError(index))
for index in range(result.NPar())])
return result
later from outside this function is called
fitresult = do_the_fit(stuff)
fitresult.Print('V')
resulting in some example output like the following
MINOS uncert actually performed? True
fit result paramters
[260129.67509556608, -19.7481953157707, 49.77154819522133, -62.054378336095624, 27.899695549113233, 0.9721094244071828]
lower, uppper MINOS uncert
[(3070.5583880112536, 3070.5583880112536), (0.04345806522217188, 0.04345806522217188), (0.07612601734777513, 0.07612601734777513), (0.12239167559851438, 0.12239167559851438), (0.1169098125378838, 0.1169098125378838), (-0.16285427481497156, 0.2619734059918487)]
****************************************
Minimizer is Minuit2 / Minimize
MinFCN = 31.7627
NDf = 0
Edm = 2.92165e-06
NCalls = 286
Par_0 = 260130 +/- 3070.56 (limited)
Par_1 = -19.7482 +/- 0.0434581 (limited)
Par_2 = 49.7715 +/- 0.076126 (limited)
Par_3 = -62.0544 +/- 0.122392 (limited)
Par_4 = 27.8997 +/- 0.11691 (limited)
Par_5 = 0.972109 +/- 0.151311 (limited)
Covariance Matrix:
Par_0 Par_1 Par_2 Par_3 Par_4 Par_5
Par_0 9.4865e+06 -104.02 25.331 78.547 -10.698 -200.82
Par_1 -104.02 0.0018887 -0.0019669 -0.00098339 0.0015277 0.00016225
Par_2 25.331 -0.0019669 0.0057954 -0.003951 -0.00078522 0.0017815
Par_3 78.547 -0.00098339 -0.003951 0.014977 -0.011507 0.0012777
Par_4 -10.698 0.0015277 -0.00078522 -0.011507 0.013666 -0.0031049
Par_5 -200.82 0.00016225 0.0017815 0.0012777 -0.0031049 0.022896
Correlation Matrix:
Par_0 Par_1 Par_2 Par_3 Par_4 Par_5
Par_0 1 -0.77707 0.10803 0.20838 -0.029711 -0.43091
Par_1 -0.77707 1 -0.59451 -0.18489 0.30069 0.024672
Par_2 0.10803 -0.59451 1 -0.42408 -0.088231 0.15465
Par_3 0.20838 -0.18489 -0.42408 1 -0.80432 0.069
Par_4 -0.029711 0.30069 -0.088231 -0.80432 1 -0.17552
Par_5 -0.43091 0.024672 0.15465 0.069 -0.17552 1
It can be seen that symmetric uncertainties were found for the first 5 parameters and MINOS uncertainties were found for the 6th parameter. What has happened such that MINOS uncertainties were not found for all 6?