Calibration of evolutionary models with branching trajectories

Develop calibration methodologies for evolutionary models that exhibit multiple possible trajectories such that predictions and tests remain meaningful; specifically, devise principled methods to determine whether unobserved outcomes were possible but not realized or were precluded by structural features of the model, thereby resolving ambiguity in model validation.

Background

The authors highlight a key methodological challenge for predictive modeling of evolutionary systems: when models produce many possible outcomes but data are sparse, straightforward calibration can either suppress genuine possibilities or fail to detect structural mis-specifications.

They emphasize that without a way to decide whether unobserved outcomes were possible, predictions become ambiguous and standard scientific validation practices are undermined.