Analysing gravitational-wave signals below conventional significance thresholds

The standard approach to publishing gravitational-wave events is deliberately conservative: only signals that pass a high significance threshold — typically a false alarm rate of one event per century or better — appear in the major catalogues produced by the LIGO, Virgo, and KAGRA collaborations. This is appropriate for ensuring catalogue reliability, but it leaves a potentially large population of real signals only partially characterised.

As the detectors have grown more sensitive, the number of marginal event candidates has grown alongside the confirmed events. Several groups outside the LVK collaborations have published their own catalogues identifying these candidates using different search methods and significance criteria. Some of these sub-threshold signals may be genuine gravitational-wave events; others may be noise artefacts. Careful analysis is needed to distinguish between them.

The subthreshold project takes two complementary approaches. The first applies LVK-quality parameter estimation methods to the candidates identified by non-LVK groups, using open data and reproducible analysis pipelines built on asimov. This produces results that can be directly compared with LVK catalogue analyses, and flags cases where different methodologies yield significantly different inferred parameters — which can point to issues with either the signal model or the analysis approach.

The second approach develops methods for coherently combining information from multiple marginal events. Individually weak signals carry limited information, but taken together they can collectively provide evidence for a real astrophysical source class — much as population-level analyses of the confirmed catalogues have revealed structure in the black hole mass and spin distribution that no single event could establish on its own.