Procoli: Profiles of cosmological likelihoods

Frequentist profile likelihoods have seen a resurgence in cosmology, offering an alternative to Bayesian methods as they can circumvent the impact of prior-volume effects. This paper presents Procoli, a fast and accessible package to obtain profile likelihoods in cosmology, available on GitHub and P...

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2024-01
Main Authors: Karwal, Tanvi, Patel, Yashvi, Bartlett, Alexa, Poulin, Vivian, Smith, Tristan L, Pfeffer, Daniel N
Format: Article
Language:English
Subjects:
Bayesian analysis
Cosmology
Dark energy
Optimization techniques
Parameter robustness
Simulated annealing
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Frequentist profile likelihoods have seen a resurgence in cosmology, offering an alternative to Bayesian methods as they can circumvent the impact of prior-volume effects. This paper presents Procoli, a fast and accessible package to obtain profile likelihoods in cosmology, available on GitHub and PyPI. Procoli seamlessly integrates with MontePython, incorporating all its available data likelihoods, as well as any modified versions of CLASS. This paper provides a comprehensive overview of the Procoli code, detailing the simulated-annealing optimizer at its core and the sequential computation of the profile. An an example, we use the early dark energy model which is afflicted by prior-volume effects to illustrate the code's features. We validate its optimizer with mock data, and compare optimization techniques for both the global minimum and the profile. Procoli further enables splitting profiles into their component contributions from individual experiments, offering nuanced insights into the data and model. As a valuable addition to the cosmologist's toolkit, Procoli supplements existing Bayesian codes, contributing to more robust parameter constraints in cosmological studies.
ISSN:2331-8422