Testing evolution models against population-level data

Develop principled statistical and machine learning approaches to rigorously test theoretical models of planetary system evolution against population-level observational datasets, enabling robust inference even when models and data are well specified.

Background

Despite progress in modeling (e.g., N-body dynamics) and abundant exoplanet demographics, drawing statistically robust conclusions remains challenging. The author points to the need for methods that bridge theory and large datasets, improving the rigor and power of population-level tests of formation and evolution scenarios.

References

Open questions and problems that I would like to see resolved include, in no particular order: How can theoretical models of planetary system evolution be tested against population-level data? Even when the theoretical model is well-specified and accurately calculable (e.g. pure N-body dynamics), and the data well-understood, inferring principled statistical constraints is often difficult. New statistical and Machine Learning approaches show promise.

— Planet formation theory: an overview  (2412.11064 - Armitage, 2024) in Section “Some open questions”, Item 8

A similar approach could be used to model the 2:1 component but we leave that for future work. The statistics will need to be improved for further progress.

— ExoMOD II. A Statistical Model of Transit Timing Variations in Kepler Multi-Planet Systems  (2608.28550 - Nesvorny et al., 28 Aug 2026) in Section 3.3, paragraph discussing Figure 9