All Models Are Wrong: Noise Modelling in Pulsar Timing

Abstract:

Pulsar timing arrays, like any experiment, depend on accurate noise modelling to detect nanohertz gravitational waves. But the noise itself is much more complex than in some other experiments. It is multi-component and, due to the astrophysical origin of several noise processes, often uncertain and difficult to model accurately. In this talk, I explore what happens when our noise models are wrong.

I show that misspecified noise can bias gravitational-wave inference and distort spectral estimates. While these effects are sometimes conservative, they can also lead to systematically incorrect conclusions, highlighting the need to account for model uncertainty properly.

I then present a transdimensional Bayesian approach for noise analysis that lets the data decide which noise processes are present in the signal. By jointly performing model selection and parameter estimation, this framework provides a more robust and statistically consistent way to analyse pulsar timing data.

Please note that the event timezone is AEDT (UTC+11 hrs)

Location

Organiser