Assessing gravitational-wave binary black hole candidates with Bayesian odds

Gravitational waves from the coalescence of binary black holes can be distinguished from noise transients in a detector network through Bayesian model selection by exploiting the coherence of the signal across the network. We present a Bayesian framework for calculating the posterior probability that a signal is of astrophysical origin, agnostic to the specific search strategy, pipeline or search domain with which a candidate is identified. We apply this framework under identical assumptions to all events reported in the LIGO-Virgo GWTC-1 catalog, GW190412 and numerous event candidates reported by independent search pipelines by other authors. With the exception of GW170818, we find that all GWTC-1 candidates, and GW190412, have odds overwhelmingly in favor of the astrophysical hypothesis, including GW170729, which was assigned significantly different astrophysical probabilities by the different search pipelines used in GWTC-1. GW170818 is de facto a single detector trigger, and is therefore of no surprise that it is disfavored as being produced by an astrophysical source in our framework. We find three additional event candidates, GW170121, GW170425 and GW170727, that have significant support for the astrophysical hypothesis, with a probability that the signal is of astrophysical origin of 0.53, 0.74 and 0.64, respectively. We carry out a hierarchical population study which includes these three events in addition to those reported in GWTC-1, finding that the main astrophysical results are unaffected.