Hierarchical_BeyondvacGR

Decoupling local and global Beyond vacuum-GR effects in EMRIs with hierarchical modeling. Used for the analysis in this paper: arxiv/2510.17398

Installation Guide

Dependencies:

• FastEMRIWaveforms(FEW)v2.0. This branch is fully compatible with the PaperProduction tag, although this issue needs fixing.
• StableEMRIFisher(SEF). This branch is fully compatible with the sef_v2 branch.
• LISAanalysistools(LAT). LISA response util.
• lisa-on-gpu. LISA response util.

The main hieararchical Class is hosted inside the hierarchical folder. Here's a brief description of the different files:

• hierarchical/Hierarchical_Class.py: the main script file hosting the Hierarchical class. Given a cosmology/EMRI parametrization setup, this script calculates (i) the Fisher matrices at the true parameter point, (ii) the biased-inference points in the vacuum, local-only, and global-only hypotheses, and (iii) Bayes factors comparing the three hypothesis.
• hierarchical/FisherValidation.py: script file that can validate Fishers calculated at the true parameter point and the biased-inference point using KL-divergence. If KL is above a given threshold, the Fishers at the two points are assumed to be 'too different', breaking the linear-signal approximation. Validation can be performed from the Hierarchical class in Hierarchical_class.py.
• hierarchical/utility.py: various utility functions (cosmology, analysis) used by Hierarchical_Class.py and FisherValidation.py.

To reproduce the results from the paper, download the datafiles here and unzip contents.

Additionally, the execution folder contains all files used to generate the populations and perform the analysis. Here's a brief description of the different files:

• execution/class_execution_Hierarchical.py: execution file with fixed f and Gdot in the true population (single evaluation).
• execution/class_execution_varied_f.py: execution file for trend where the fractional EMRIs with a local effect, f, varies from 0 to 1. Gdot (global effect amplitude) is fixed to 1e-12.
• execution/class_execution_varied_Gdot.py: execution file where the global effect amplitude (Gdot) varies. f is fixed to 0.5.

Citations

@article{Kejriwal:2025jao,
    author = "Kejriwal, Shubham and Barausse, Enrico and Chua, Alvin J. K.",
    title = "{Hierarchical modeling of gravitational-wave populations for disentangling environmental and modified-gravity effects}",
    eprint = "2510.17398",
    archivePrefix = "arXiv",
    primaryClass = "gr-qc",
    month = "10",
    year = "2025"
}