New Statistical Techniques in the Measurement of the inclusive Top Pair Production Cross Section

TL;DR

This paper explores various multivariate statistical techniques, including machine learning models and homogeneity tests, to improve the measurement of the top quark pair production cross section in proton-antiproton collisions at Fermilab.

Contribution

It introduces the application and comparison of multiple advanced statistical and machine learning methods for signal-background separation in top quark pair production analysis.

Findings

Neural Networks with Switching Units perform well in signal discrimination.

Homogeneity tests identify variables with good data-MC agreement.

Comparison shows different methods vary in discrimination effectiveness.

Abstract

We present several different types of multivariate statistical techniques used in the measurement of the inclusive top pair production cross section in $p \bar{p}$-collisions at $\sqrt{s} = 1.96 \text{TeV}$ employing the full RunII data ($9.7\textrm{ fb}^{-1}$) collected with the D0 detector at the Fermilab Tevatron Collider. We consider the final state of the top quark pair decays containing one electron or muon and at least two jets. We proceed various statistical homogeneity tests such as Anderson - Darling, Kolmogorov - Smirnov, and $\varphi$-divergences tests to determine, which variables have good data-MC agreement, as well as a good separation power. We adjusted all tests for using weighted empirical distribution functions. Further we separate $t\bar{t}$ signal from the background by the application of Generalized Linear Models, Gaussian Mixture Models, Neural Networks with Switching Units and confront them with familiar methods from ROOT TMVA package such as Boosted Decision Trees, and Multi-layer Perceptron. We compare results by area under receiver operating characteristic curve and verify the quality of the discrimination from all methods.