Debiasing standard siren inference of the Hubble constant with marginal neural ratio estimation

Abstract

Gravitational wave (GW) standard sirens may resolve the Hubble tension, provided that standard siren inference of H0 is free from systematic biases. However, standard sirens from binary neutron star (BNS) mergers suffer from two sources of systematic bias, one arising from the anisotropy of GW emission, and the other from the anisotropy of electromagnetic (EM) emission from the kilonova. For an observed sample of BNS mergers, the traditional Bayesian approach to debiasing involves the direct computation of the detection likelihood. This is infeasible for large samples of detected BNS merger due to the high dimensionality of the parameter space governing merger detection. In this study, we bypass this computation by fitting the Hubble constant to forward simulations of the observed GW and EM data under a simulation-based inference (SBI) framework using marginal neural ratio estimation. A key innovation of our method is the inclusion of BNS mergers which were only detected in GW, which allows for estimation of the bias introduced by EM anisotropy. Our method corrects for $\sim 90\%$ of the bias in the inferred value of H0 when telescope follow-up observations of BNS mergers have extensive tiling of the merger localization region, using known telescope sensitivities and assuming a model of kilonova emission. Our SBI-based method thus enables a debiased inference of the Hubble constant of BNS mergers, including both mergers with detected EM counterparts and those without.

Cosmography with next-generation gravitational wave detectors

“…Over the last few years, the systematic uncertainties associated with bright sirens have been widely studied, such as the systematics due to the GW instrumental calibration uncertainties [74][75][76], the EM observation selection effect [77][78][79], the biased EM-inferred binary viewing angle [77], the peculiar velocity of the hosts [80][81][82], and the GW instrumental nonstationary noise [83][84][85][86][87]. A comprehensive study of the systematics and the developments of the mitigation methods are critical to ensure the value of standard siren measurements in cosmology, especially when dealing with the large and precise catalog of bright sirens that XG detectors will provide.…”

A thorough investigation of the prospects of eLISA in addressing the Hubble tension: Fisher forecast, MCMC and Machine Learning

“…This entails the need for faster and more efficient computational tools and data handling algorithms. Besides conventional methods of simulation and data analysis, various machine learning (ML) techniques like Gaussian Processes (GP), Genetic Algorithms (GA), and various deep learning algorithms are increasingly being used in different areas of cosmology (for a small body of diverse examples from recent years see [60][61][62][63][64][65][66][67][68][69][70][71][72][73][74][75][76][77][78]). Gaussian Processes, for example, have already found considerable application in the area of non-parametric reconstructions of various cosmological parameters [79][80][81][82].…”

EFTofLSS meets simulation-based inference: σ 8 from biased tracers

“…In the context of galaxy clustering, recent progress has been made by using graphs networks to capture the map information from the galaxy distribution [22,23]. SBI has been widely used in the cosmological inference context, including weak-lensing [24][25][26][27], type IA supernovae [24,25,[28][29][30][31][32], standard sirens [33,34], CMB [35,36], galaxy cluster abundance [37], Gaussian and lognormal fields [38][39][40][41], dark-matter overdensity fields [40,42], voids [43], dark-matter halos [23,44] and galaxies [45][46][47][48].…”