A fast and stable approximate maximum-likelihood method for template fits

TL;DR

This paper introduces a new approximate maximum-likelihood method for template fitting that is faster and more stable, especially with weighted data and large bin numbers, while maintaining comparable accuracy to the exact likelihood.

Contribution

A novel approximate likelihood derived from the exact Barlow-Beeston likelihood, generalized for weighted data, offering improved speed and stability in template fitting.

Findings

Performance comparable to the exact likelihood in unweighted cases.

Better performance in weighted template fits.

Faster evaluation with increasing number of bins.

Abstract

Barlow and Beeston presented an exact likelihood for the problem of fitting a composite model consisting of binned templates obtained from Monte-Carlo simulation which are fitted to equally binned data. Solving the exact likelihood is technically challenging, and therefore Conway proposed an approximate likelihood to address these challenges. In this paper, a new approximate likelihood is derived from the exact Barlow-Beeston one. The new approximate likelihood and Conway's likelihood are generalized to problems of fitting weighted data with weighted templates. The performance of estimates obtained with all three likelihoods is studied on two toy examples: a simple one and a challenging one. The performance of the approximate likelihoods is comparable to the exact Barlow-Beeston likelihood, while the performance in fits with weighted templates is better. The approximate likelihoods evaluate faster than the Barlow-Beeston one when the number of bins is large.