Calibrating signal-to-noise ratio detection thresholds using gravitational-wave catalogs

Searching for gravitational-wave signals is a challenging and computationally intensive endeavor undertaken by multiple independent analysis pipelines. While detection depends only on observed noisy data, it is sometimes inconsistently defined in terms of source parameters that in reality are unknown, e.g., by placing a threshold on the optimal signal-to-noise ratio (SNR). We present a method to calibrate unphysical thresholds to search results by performing Bayesian inference on real observations using a model that simultaneously parametrizes the intrinsic network optimal SNR distribution and the effect of search sensitivity on it. We find consistency with a fourth-order power law and detection thresholds of 10.5-2.4+2.1, 11.2-1.4+1.2, and 9.1-0.5+0.5 (medians and 90% credible intervals) for events with false-alarm rates less than 1 yr-1 in the first, second, and third LIGO-Virgo-KAGRA observing runs, respectively. Though event selection can only be self-consistently reproduced by physical searches, employing our inferred thresholds allows approximate observation-calibrated selection criteria to be applied when efficiency is required and injection campaigns are infeasible.