A Coincidence Null Test for Poisson-Distributed Events

Abstract

When transient events are observed with multiple sensors, it is often necessary to establish the significance of coincident events. We derive a universal null test for an arbitrary number of sensors motivated by the archetypal detection problem for independent Poisson-distributed events in gravitational-wave detectors such as LIGO and Virgo. In these detectors, transient events may be witnessed by myriad channels that record interferometric signals and the surrounding physical environment. We apply our null test to a broad set of simulated gravitational-wave events as well as to a real gravitational-wave detection to determine which auxiliary channels do and do not witness real gravitational waves, and therefore which are safe to use when constructing vetoes. We also describe how our approach can be used to study detector artifacts and their origin, as well as to quantify the statistical independence of candidate GW signals from noise artifacts observed in auxiliary channels.

LIGO detector characterization in the second and third observing runs

“…Algorithms such as the pointy statistic [73] and HVeto then run analyses on the hardware injections to generate lists of safe and unsafe channels. pointy is a null-test that uses the assumption that events in auxiliary channels are distributed according to stationary Poisson processes.…”