Hierarchical Bayesian Models for Type Ia Supernova Inference

Abstract

Type Ia supernovae (SN Ia) are the most precise cosmological distance indicators and are important for measuring the acceleration of the Universe and the properties of dark energy. Current cosmological analyses use rest-frame optical SN Ia light curves to estimate distances, whose accuracy is limited by the confounding effects of host galaxy dust extinction. The combination of broadband optical and near-infrared (NIR) light curves and spectroscopic data has the potential to improve inference in supernova cosmology. I describe a principled, hierarchical Bayesian framework to coherently model the multiple random and uncertain effects underlying the observed data, including measurement error, intrinsic supernova covariances, host galaxy dust extinction and reddening, peculiar velocities and distances. Using a new MCMC code, BayeSN, to compute probabilistic inferences for individual SN Ia and the population, I applied these hierarchical models to the joint analysis of the optical, near-infrared (NIR), and spectroscopic data from a large sample of nearby SN Ia. The combination of optical and NIR data better constrains estimates of the dust effects and approximately doubles the precision of cross-validated SN Ia distance predictions compared to using optical data alone. The hierarchical model is extended to include spectroscopic data to estimate significant correlations between the intrinsic optical colors and ejecta velocities. These applications demonstrate the power and flexibility of multi-level modeling in the analysis of SN Ia data.