A solution for performing maximum likelihood estimation on models built from histogram templates.
- Free software: GNU General Public License v3
- Documentation: https://histimator.readthedocs.io.
the histimator core directory has a file called Models containing the core HistiModel class.
the model is initialised:
from histimator.models import HistiModel
model = HistiModel("model name")
Each channel is defined as:
from histimodel.Channel import HistiChannel
SR = HistiChannel("SignalRegion")
data can be added to the channels as:
SR.SetData([list of data points])
any number of samples are defined as:
from histimator.models import HistiSample
sig = HistiSample("Signal")
bkg = HistiSample("Background")
each of which needs a histogram:
sig.SetHisto(numpy.histogram) bkg.SetHisto(numpy.histogram)
currently the only parameters available are an overal normalisation on these templates. this is given with a name an initial value (default 1) and a range (default [0.1,10]). Currently no implementation is actually in place to tell Minuit about this range...:
sig.AddNorm("some_norm",1,0,3)
Finally, the samples must be added to the channel and this added to the model.:
SR.AddSample(sig) SR.AddSample(bkg) model.AddChannel(SR)
This model can now be evaluated using probfit Binned Likelihood function:
from iminuit import Minuit from probfit import BinnedLH blh = BinnedLH(model.pdf, data, bins=10, bound=bound, extended=True) m = Minuit(blh, some_norm=0.5, error_some_norm=1.5) m.migrad()
this has various built in plotting functionality.
