A Bayesian coincidence test for noise rejection in a gravitational-wave burst search

In searches for gravitational-wave bursts, a standard technique used to reject noise is to discard burst event candidates that are not seen in coincidence in multiple detectors. A coincidence test in which Bayesian inference is used to measure how noise-like a tuple of events appears is presented here. This technique is shown to yield higher detection efficiencies for a given false alarm rate than do techniques based on per-parameter thresholds when applied to a toy model covering a broad class of event candidate populations. Also presented is the real-world example of a use of the technique for noise rejection in a time–frequency burst search conducted on simulated gravitational-wave detector data. Besides achieving a higher detection efficiency, the technique is significantly less challenging to implement well than is a per-parameter threshold method.