Data analysis strategies for the detection of gravitational waves in non-Gaussian noise

TL;DR

This paper develops a robust statistical detection method for gravitational waves in non-Gaussian, non-stationary noise, introducing a two-component noise model and comparing detection strategies.

Contribution

It presents a new two-component noise model and derives an optimal detection statistic that effectively suppresses noise bursts in gravitational wave data analysis.

Findings

The optimal detection statistic includes a natural veto for noise bursts.

A simple coincidence detection approximates the optimal statistic in two-detector scenarios.

The proposed methods outperform standard Gaussian-based detection in simulated non-Gaussian noise.

Abstract

In order to analyze data produced by the kilometer-scale gravitational wave detectors that will begin operation early next century, one needs to develop robust statistical tools capable of extracting weak signals from the detector noise. This noise will likely have non-stationary and non-Gaussian components. To facilitate the construction of robust detection techniques, I present a simple two-component noise model that consists of a background of Gaussian noise as well as stochastic noise bursts. The optimal detection statistic obtained for such a noise model incorporates a natural veto which suppresses spurious events that would be caused by the noise bursts. When two detectors are present, I show that the optimal statistic for the non-Gaussian noise model can be approximated by a simple coincidence detection strategy. For simulated detector noise containing noise bursts, I compare the operating characteristics of (i) a locally optimal detection statistic (which has nearly-optimal behavior for small signal amplitudes) for the non-Gaussian noise model, (ii) a standard coincidence-style detection strategy, and (iii) the optimal statistic for Gaussian noise.