Bayesian Nested Sampling for Robust Model Selection and Validation of Lunar Occultation Light Curves Agus Triono Puri Jatmiko
Bosscha Observatory, ITB
Abstract
Bayesian Nested Sampling (NS) offers a powerful framework for model selection and validation in astrophysical data analysis. This study applies NS to analyze lunar occultation (LO) light curves, enabling robust sampling of the posterior distribution for model selection and validation purposes. However, model selection for these light curves is challenging due to complex noise structures and parameter degeneracies. We implement a Bayesian approach to model the LO light curves, incorporating multi-level priors to account for uncertainties in noise models and physical parameters. NS is employed to compute the Bayesian evidence, facilitating the comparison of competing models, such as single-star, binary, or multiple-stars configurations. Our methodology leverages NS^s ability to efficiently explore high-dimensional parameter spaces and marginalize over nuisance parameters, ensuring accurate model ranking and parameter estimation. We validate our approach using synthetic LO light curves generated from known stellar configurations. Our next step is incorporating this method as part of our lunar occultation pipeline to ensure the proper selection of the model.
