One to many: comparing single gravitational-wave events to astrophysical populations

Gravitational-wave observations have revealed sources whose unusual properties challenge our understanding of compact-binary formation. Inferring the formation processes that are best able to reproduce such events may therefore yield key astrophysical insights. A common approach is to count the fraction of synthetic events from a simulated population that are consistent with some real event. Though appealing owing to its simplicity, this approach is flawed because it neglects the full posterior information, depends on an ad-hoc region that defines consistency, and fails for high signal-to-noise detections. We point out that a statistically consistent solution is to compute the posterior odds between two simulated populations, which crucially is a relative measure, and show how to include the effect of observational biases by conditioning on source detectability. Applying the approach to several gravitational-wave events and simulated populations, we assess the degree to which we can conclude model preference not just between distinct formation pathways but also between subpopulations within a given pathway.