A multidimensional, event-by-event, statistical weighting procedure for signal to background separation

Carnegie Mellon University, Pittsburgh, Pennsylvania 15213, USA

^* e-mail: zbaldwin@cmu.edu

Published online: 6 May 2024

Abstract

Numerous analyses performed in nuclear and particle physics are in search of signals that are contaminated by irreducible background that cannot be suppressed using event-selection criteria. These background events can lead to unphysical or biased results when extracting physical observables and need to be taken into account. Exploring a data set across multiple dimensions allows us to characterize the phase space of a desired reaction through a set of coordinates. For a subset of these coordinates, known as reference coordinates, signal and background follow different distributions with known functional forms with potential unknown parameters. The Quality Factor approach uses the space defined by the remaining non-reference phase space coordinates to determine the k-nearest neighbors of an event. The distribution of these neighbors in the reference coordinates undergoes a fit with the sum of the signal and background model functions, employing techniques like the unbinned maximum likelihood method, to extract the signal fraction, or Q-factor. This quality factor, which is defined for each event, is equal to the probability that it originates from the signal of interest. In this document, we will give a brief overview of this procedure and illustrate examples using Monte Carlo simulations and data from the GlueX experiment at Jefferson Lab.