Primary Submission Category: Applications in Health and Biology
Three-sided Testing with Two Control Groups with an Application to an Observational Study of the Impact of High School Football Participation on Subsequent Cognitive Functioning
Authors: Iris Horng, Da Wu, Sameer Deshpande, Dylan Small,
Presenting Author: Iris Horng*
American football, the most popular high school sport for American boys, has sparked concern about its long-term cognitive effects. Using the Wisconsin Longitudinal Study dataset, which tracks individuals who graduated from Wisconsin high schools in 1957, we focus on their outcomes of letter fluency and delayed word recall at ages 65 and 72. While prior studies have used binary hypothesis tests for cognitive decline, such approaches cannot determine whether an effect is small enough to be considered negligible. To address this, we use a three-sided testing framework that simultaneously assesses superiority, inferiority, and equivalence. We further extend this framework to incorporate two distinct control groups – non-contact sport participants and non-athletes — to improve robustness to hidden bias. Our main contribution is a rigorous methodology for integrating two control groups within three-sided testing, while ensuring proper control of familywise error. Assuming limited unmeasured confounding, we find no meaningful impact of playing high school football on later life cognitive functioning. An R package, sen2controlgroups, implements the method.