Reconstructing and resampling: a guide to utilising posterior samples from gravitational wave observations

The LIGO, Virgo, and KAGRA (LVK) gravitational-wave observatories have opened new scientific research in astrophysics, fundamental physics, and cosmology. The collaborations that build and operate these observatories release the interferometric strain data as well as a catalogue of observed signals with accompanying Bayesian posterior distributions. These posteriors, in the form of equally-weighted samples, form a dataset that is used by a multitude of further analyses seeking to constrain the population of merging black holes, identify lensed pairs of signals, and much more. However, many of these analyses rely, often implicitly, on the ability to reconstruct the likelihood and prior from the inputs to the analysis and apply resampling (a statistical technique to generate new samples varying the underlying analysis assumptions). In this work, we first provide a guide on how to reconstruct and modify the posterior density accurately from the inputs for analyses performed with the Bilby inference library. We then demonstrate and compare resampling techniques to produce new posterior sample sets and discuss Pareto-smoothing to improve the efficiency. Finally, we provide examples of how to use resampling to study observed gravitational-wave signals. We hope this guide provides a useful resource for those wishing to use open data products from the LVK for gravitational-wave astronomy.