Gravitational-wave hierarchical inference from population samples

When doing gravitational-wave population fits (the notorious Bayesian hierarchical analyses), one typically relies on models that can be evaluated. That means I have the PDF of the model. This seems fair—what’s wrong with that? But in practice, it prevents a lot of applications. My stellar physics friends who run state-of-the-art simulations don’t really have a PDF, but rather a numerical generative model. In practice, they can give me a table of simulated sources; that is, they have “samples” from a PDF, but the PDF itself is not easily evaluable. Great work by Federico here, who in this paper has rewritten the GW population problem in a way that can make use of these samples, rather than PDF evaluations. This opens the door to so many applications!

F. De Santi, D. Gerosa, M. Mould.
arXiv:2611.17428 [[astro-ph.HE]].