Frequentist limit setting in effective field theories

TL;DR

This paper examines frequentist methods for setting confidence intervals in effective field theories, highlighting issues with alternative approaches that may over-constrain parameters when the signal depends quadratically on the parameter.

Contribution

It identifies potential problems with common frequentist interval methods in quadratic signal scenarios, emphasizing the robustness of the Feldman-Cousins approach.

Findings

Alternative methods can over-constrain parameters in quadratic signal cases

The original Feldman-Cousins method avoids encoding goodness-of-fit into intervals

Problems arise when applying these methods to effective theories with quadratic dependence

Abstract

The original frequentist approach for computing confidence intervals involves the construction of the confidence belt which provides a mapping of the observation in data into a subset of values for the parameter. There are different prescriptions for constructing the confidence belt, here we use the one provided by Feldman and Cousins. Alternative methods based on the frequentist idea exist, including the delta likelihood method, the $CL_s$ method and a method here referred to as the $p$-value method, which have all been commonly used in high energy experiments. The purpose of this article is to draw attention to a series of potential problems when applying these alternative methods to the important case where the predicted signal depends quadratically on the parameter of interest, a situation which is common in high energy physics as it covers scenarios encountered in effective theories. These include anomalous Higgs couplings and anomalous trilinear and quartic gauge couplings. It is found that the alternative methods, contrary to the original method using the confidence belt, encode the goodness-of-fit into the confidence intervals and potentially over-constrain the parameter.