A generalized likelihood ratio test statistic for Cherenkov telescope data

TL;DR

This paper introduces a new, robust likelihood ratio test statistic tailored for Cherenkov telescope data, especially effective in wobble mode and irregular acceptance scenarios, enhancing detection and skymapping capabilities.

Contribution

It proposes a generalized likelihood ratio test statistic that accounts for multi-pointing observations and instrument acceptance variations, improving analysis robustness.

Findings

The new test statistic performs well in Monte Carlo simulations.

It is effective for source detection, skymapping, and shape fitting.

The method handles irregular acceptance and wobble mode data.

Abstract

Astrophysical sources of TeV gamma rays are usually established by Cherenkov telescope observations. These counting type instruments have a field of view of few degrees in diameter and record large numbers of particle air showers via their Cherenkov radiation in the atmosphere. The showers are either induced by gamma rays or diffuse cosmic ray background. The commonly used test statistic to evaluate a possible gamma-ray excess is Li and Ma (1983), Eq. 17, which can be applied to independent on- and off-source observations, or scenarios that can be approximated as such. This formula however is unsuitable if the data are taken in so-called "wobble" mode (pointing to several offset positions around the source), if at the same time the acceptance shape is irregular or even depends on operating parameters such as the pointing direction or telescope multiplicity. To provide a robust test statistic in such cases, this paper explores a possible generalization of the likelihood ratio concept on which the formula of Li and Ma is based. In doing so, the multi-pointing nature of the data and the typically known instrument point spread function are fully exploited to derive a new, semi-numerical test statistic. Due to its flexibility and robustness against systematic uncertainties, it is not only useful for detection purposes, but also for skymapping and source shape fitting. Simplified Monte Carlo simulations are presented to verify the results, and several applications and further generalizations of the concept are discussed.