GWTC-4 Multi Spin: Population Inference Enhanced by Analytical Likelihood Models

The growing catalog of gravitational wave events has revealed a rich diversity in the properties

of compact binary mergers. However, the current population inference methods often struggle to

constrain the properties of narrow populations, leading to unreliable estimates of population param-

eters. In this study, we assessed the impact of using sample based and continuous approximations

on narrow population inferences. We simulated populations of binary black holes with power law

mass distribution and gaussian spin component distribution centered at zero, and performed pop-

ulation inference using Bayesian parametric models implemented in GWKokab. The continuous

approximation directly used a gaussian likelihood function instead of sampling priors like in the

discrete method. The continuous approximation, unlike the sample based method, better recovered

the simulated narrow population. The use of analytical likelihood models can significantly enhance

the accuracy and efficiency of population inference and probe real astrophysical data to ascertain

the true population of objects in the universe. Furthermore, to explore the potential of this method

in real data, we applied it to the LIGO-Virgo-KAGRA (LVK) GWTC-4 catalog by building multi-spin model which has independent spins to broken power law and gaussian compoennts.