Examining the Research Designs and Analytical Techniques in Athletic Training from 2005-2010
In May 2010, the editorial board for the Journal of Athletic Training (JAT), the top rated journal in athletic training, published an editorial calling for a change in the terminology of the design section of future manuscript submissions. The intent was to align the design section of articles in the JAT with other journals in medicine. The editorial board identified 13 design categories, and while this is a critical first step in identifying the research design, I would argue that further detail should be included in the design section. This is critical since the appropriate analyses hinge on the research design. In athletic training, longitudinal designs are very common. I reviewed the published articles in the JAT from 2005–2010 and found the most common method to analyze data from longitudinal designs is a repeated measures analysis of variance (RM ANOVA). Numerous assumptions must be met for the results from an RM ANOVA to be valid. It is common that these assumptions are not met in longitudinal designs. The purpose of this dissertation is to present an alternative statistical method for data from longitudinal designs, compared to the traditional RM ANOVA. I propose using a hierarchical linear model (HLM), which has been more commonly used, with success, in other medical fields. Through a purposeful random sampling, 18 corresponding authors with articles in the JAT from 2005–2010 that had a longitudinal design and used an RM ANOVA were contacted through email. Nine authors were willing to provide the de-identified data presented in the journal article. I replicated the 2-way RM ANOVA described in the published article and compared the findings to the published results. Only 2 articles mentioned assumption testing. Yet my reanalysis revealed that the assumption of sphericity was violated in all datasets. I then used an HLM to reexamine the same data. The HLM analysis focuses on modeling individual growth trajectories and answers different questions than the RM ANOVA. However, practical implications can be drawn from either analysis. The practical implications were similar using either HLM or RM ANOVA in 4 cases and differed in 4 cases.