With close to 100 confident observations, the field of gravitational-wave astronomy has been firmly established, enabling new tests of general relativity and independent calculation of astronomical results such as the Hubble constant. Bayesian inference under- pins most analytical methodologies of gravitational waves, allowing scientists to extract unknown parameters from gravitational-wave signals. Implementations of Bayesian methodologies typically model strain data as stationary, although this is not representative of LIGO or Virgo detector noise. In this thesis, we demonstrate that the posterior of estimated waveform parameters is no longer valid under the assumption of stationarity. Although this does not have the effect of biasing estimated parameters, they will be under- or overestimated compared to the true posterior values. This effect will grow more significant the longer the signal. We demonstrate how we incorporated the non-stationary noise model into inference methodologies, by approximating the non-stationarity as a low-dimensional component of the noise covariance. We also show how to approximate the latent non-stationary modulation of LIGO strain data, to accurately approximate the noise covariance for realistic noise. We expect to observe signals of duration on the order of minutes or longer in future detectors, such as LISA or the Einstein Telescope, and so the assumption of stationarity will completely break down and estimated parameters will not be accurately measured. Accounting for this limiting factor in parameter sensitivity could be very important for achieving accurate astronomical results. Therefore, a methodology such as has been described here will be vital for future analyses.
Date of Award | 21 Feb 2024 |
---|
Original language | English |
---|
Awarding Institution | |
---|
Supervisor | Andrew Lundgren (Supervisor) & David Bacon (Supervisor) |
---|
Non-Stationarity in Gravitational-Wave Analysis
Edy, O. P. (Author). 21 Feb 2024
Student thesis: Doctoral Thesis