Statistical power analysis in exercise science

The replicability crisis has widespread debate in many academic disciplines, yet it remains an underexplored issue in exercise science. A major factor contributing to the lack of replicability across disciplines is the low statistical power of studies. In contrast to the common interpretation of low statistical power, where low power only decreases the probability of finding an effect, studies with insufficient statistical power are prone to misinterpreting the true effects of their findings, undermining the integrity of scientific evidence. This exploratory study aims to assess the typical statistical power in exercise science research. This review includes 90 meta-analyses published in the last five years in exercise science, containing N = 1191 individual studies. The statistical power of each study was calculated using the aggregated effect sizes reported in the meta-analyses as an approximation of true effect sizes. The results indicate that the median statistical power of the studies reviewed may be as low as 30%. This result should be rigorously tested in future confirmatory studies.