Setting confidence intervals in coincidence search analysis

Abstract

The main technique that has been used to estimate the rate of gravitational wave (gw) bursts is to search for coincidence among times of arrival of candidate events in different detectors. Coincidences are modeled as a (possibly non-stationary) random time series background with gw events embedded in it, at random times but constant average rate. It is critical to test whether the statistics of the coincidence counts is Poisson, because the counts in a single detector often are not. At some point a number of parameters are tuned to increase the chance of detection by reducing the expected background: source direction, epoch vetoes based on sensitivity, goodness-of-fit thresholds, etc. Therefore, the significance of the confidence intervals itself has to be renormalized. This review is an insight of the state-of-the-art methods employed in the recent search performed by the International Gravitational Event Collaboration for the worldwide network of resonant bar detectors.