Quantile methods for longitudinal running data: A New York City Marathon example

Increased participation and data collection in long-distance running events resulted in many large datasets featuring intermediate time recording. Yet, often they are not explored to their full potential. Existing research frequently performs analyses on summary statistics only, and often selects one performance group. While their typical ANOVA analysis is useful for mean-based questions, complementary (distributional) information can be obtained by applying quantile tools, also enabling a meaningful study of both amateur and elite runners in a unified framework. The contributions of the present article are threefold. (i) Recent quantile methodology for longitudinal data, including quantile-conditioned curves and longitudinal quantile curves is introduced in an accessible way. (ii) Novel, quantile-based analyses of data on almost 47,000 finishers of the 2022 New York City Marathon are provided, illustrating how both general understanding in pacing behaviour as well as answers to more specific, application-driven research questions can be obtained. The quantile tools are consistent with existing results and additionally yield new insights. (iii) Other possible domains of application are discussed. These three aspects combined serve the purpose of facilitating the dissemination of practical takeaways for the data at hand, and the quantile-based methodology itself, to a broader audience facing longitudinal sports data.