A1606
Title: Hierarchical Bayesian joint modeling of athlete performance trajectories and latent peak ages
Authors: Kangyi Peng - Simon Fraser University (Canada) [presenting]
Tianyu Guan - York University (Canada)
Xiong Zhao - University of Ottawa (Canada)
Abstract: The focus in on the fundamental problem of determining when athletes reach peak performance, a key question in competitive sports. Traditional analyses often rely on simple summary statistics or per-athlete curve fitting, which limit inference and lack principled uncertainty quantification. A Bayesian joint modeling framework is developed that characterizes athlete performance trajectories together with their latent peak ages, specifying performance curves flexibly and embedding peak age in a hierarchical survival sub-model. The framework enables information borrowing across athletes and provides uncertainty quantification even with sparse or irregular observations. Simulation studies show that the method yields stable estimation and improves uncertainty assessment compared with conventional approaches. Analysis of World Athletics data across ten sport categories reveals discipline-specific peak-age patterns and covariate effects. Beyond sports, the framework is broadly applicable to problems involving joint modeling of longitudinal trajectories and function-derived latent features.
