Robust parameter estimation from pulsar timing data

TL;DR

This paper compares various sampling algorithms for robustly estimating parameters from pulsar timing data, aiming to improve gravitational wave background detection reliability.

Contribution

It introduces two alternative samplers for pulsar timing analysis to enhance robustness and cross-validation of results.

Findings

Consistent parameter estimates across different samplers.

Proposed two new samplers for future robustness checks.

Analysis of sampling efficiency and runtime.

Abstract

Recently, global pulsar timing arrays have released results from searching for a nano-Hertz gravitational wave background signal. Although there has not been any definite evidence of the presence of such a signal in residuals of pulsar timing data yet, with more and improved data in future, a statistically significant detection is expected to be made. Stochastic algorithms are used to sample a very large parameter space to infer results from data. In this paper, we attempt to rule out effects arising from the stochasticity of the sampler in the inference process. We compare different configurations of nested samplers and the more commonly used markov chain monte carlo method to sample the pulsar timing array parameter space and account for times taken by the different samplers on same data. Although we obtain consistent results on parameters from different sampling algorithms, we propose two different samplers for robustness checks on data in the future to account for cross-checks between sampling methods as well as realistic run-times.