The Impact of Bayesian Hyperpriors on the Population-Level Eccentricity Distribution of Imaged Planets

TL;DR

This study investigates how the choice of hyperpriors in hierarchical Bayesian modeling affects the inferred population-level eccentricity distributions of directly imaged substellar companions, emphasizing the importance of hyperprior selection especially with small samples.

Contribution

The paper demonstrates the significant impact of hyperprior choices on eccentricity distribution inference and provides an open-source Python package for community use.

Findings

Hyperprior choice significantly affects eccentricity distribution results.

The inferred eccentricity distribution aligns with close-in exoplanets.

Long-period giant planets and brown dwarfs have different eccentricity distributions.

Abstract

Orbital eccentricities directly trace the formation mechanisms and dynamical histories of substellar companions. Here, we study the effect of hyperpriors on the population-level eccentricity distributions inferred for the sample of directly imaged substellar companions (brown dwarfs and cold Jupiters) from hierarchical Bayesian modeling (HBM). We find that the choice of hyperprior can have a significant impact on the population-level eccentricity distribution inferred for imaged companions, an effect that becomes more important as the sample size and orbital coverage decrease to values that mirror the existing sample. We reanalyse the current observational sample of imaged giant planets in the 5-100 AU range from Bowler et al. (2020) and find that the underlying eccentricity distribution implied by the imaged planet sample is broadly consistent with the eccentricity distribution for close-in exoplanets detected using radial velocities. Furthermore, our analysis supports the conclusion from that study that long-period giant planets and brown dwarf eccentricity distributions differ by showing that it is robust to the choice of hyperprior. We release our HBM and forward modeling code in an open-source Python package, ePop!, and make it freely available to the community.