Refinable modeling for unbinned SMEFT analyses

TL;DR

This paper develops new machine learning-based techniques for estimating systematic uncertainties in unbinned LHC data analyses, specifically for constraining SMEFT Wilson coefficients, enhancing the statistical framework for beyond standard model research.

Contribution

It introduces a novel likelihood ratio surrogate method and a tree-boosting algorithm for modeling systematic effects in unbinned analyses, applicable beyond SMEFT.

Findings

Effective likelihood ratio surrogates for unbinned data

A new tree-boosting algorithm for systematic modeling

Application to top quark pair production SMEFT analysis

Abstract

We present techniques for estimating the effects of systematic uncertainties in unbinned data analyses at the LHC. Our primary focus is constraining the Wilson coefficients in the standard model effective field theory (SMEFT), but the methodology applies to broader parametric models of phenomena beyond the standard model (BSM). We elevate the well-established procedures for binned Poisson counting experiments to the unbinned case by utilizing machine-learned surrogates of the likelihood ratio. This approach can be applied to various theoretical, modeling, and experimental uncertainties. By establishing a common statistical framework for BSM and systematic effects, we lay the groundwork for future unbinned analyses at the LHC. Additionally, we introduce a novel tree-boosting algorithm capable of learning highly accurate parameterizations of systematic effects. This algorithm extends the existing toolkit with a versatile and robust alternative. We demonstrate our approach using the example of an SMEFT interpretation of highly energetic top quark pair production in proton-proton collisions.