Application of a generalized likelihood ratio test statistic to MAGIC data

TL;DR

This paper introduces a generalized likelihood ratio test statistic tailored for MAGIC Cherenkov telescope data, improving detection sensitivity and reducing systematics compared to traditional methods like Li & Ma.

Contribution

It applies a new, more flexible likelihood ratio test to MAGIC data, demonstrating enhanced sensitivity and robustness over existing detection statistics.

Findings

The generalized LRT outperforms Li & Ma in sensitivity.

It reduces systematic uncertainties in gamma-ray detection.

Application to archival data validates the method's effectiveness.

Abstract

The commonly used detection test statistic for Cherenkov telescope data is Li & Ma (1983), Eq. 17. It evaluates the compatibility of event counts in an on-source region with those in a representative off-region. It does not exploit the typically known gamma-ray point spread function (PSF) of a system, and in practice its application requires either assumptions on the symmetry of the acceptance across the field of view, orMonte Carlo simulations.MAGIC has an azimuth-dependent, asymmetric acceptance which required a careful review of detection statistics. Besides an adapted Li & Ma based technique, the recently presented generalized LRT statistic of [1] is now in use. It is more flexible, more sensitive and less systematics-affected, because it is highly customized for multi-pointing Cherenkov telescope data with a known PSF. We present the application of this new method to archival MAGIC data and compare it to the other, Li&Ma-based method.