PDF from a histogram in zfit?

Question

I looking in the documentation how to obtain a PDF from a histogram, but I couldn't find anything, so how can I obtain PDF from a histogram ?, for example to use it in a sum_pdf = zfit.pdf.SumPDF([model1, model2], fracs=frac) in order to perfome a fit, or maybe generate some toys.

Thanks in advance.

PS. I'm looking something similar to RooHistPdf Class from Roofit.

Answer 1

Updated answer

zfit now allows to do binned fits (to be installed currently with pip install zfit --pre) as described in the tutorial

Basically, starting from the unbinned data or model, you can do:

# make binned
binning = zfit.binned.RegularBinning(50, -8, 10, name="x")
obs_bin = zfit.Space("x", binning=binning)

data = data_nobin.to_binned(obs_bin)
model = zfit.pdf.BinnedFromUnbinnedPDF(model_nobin, obs_bin)

Old answer

There is currently no out-of-the-box solution for this but work-in-progress.

However, you can simply construct something on your own like:

import zfit
from zfit import z
import numpy as np
import tensorflow as tf

zfit.settings.options['numerical_grad'] = True


class BinnedEfficiencyPDF(zfit.pdf.BasePDF):

    def __init__(self, efficiency, eff_bins, obs, name='BinnedEfficiencyPDF'):
        self.efficiency = efficiency
        self.eff_bins = eff_bins
        super().__init__(obs=obs, name=name)

    def _binContent(self, x):
        eff_bin = np.digitize(x, self.eff_bins)
        return self.efficiency[eff_bin]

    def _unnormalized_pdf(self, x):  # or even try with PDF
        x = z.unstack_x(x)
        probs =  z.py_function(func=self._binContent, inp=[x], Tout=tf.float64)
        probs.set_shape(x.shape)
        return prob