Statistical hypothesis testing is widely used across research fields; however, its misinterpretation can easily lead to misleading conclusions. This article highlights two common issues: the misconception that non-significant results imply no effect, and the inflation of Type I error rates due to multiple testing. Using the paired t-test as an example, we explain the asymmetry of hypothesis testing, the properties of p-values, and the usefulness of confidence intervals when interpreting non-significant results. Regarding multiple testing, we introduce the Bonferroni correction as a widely used method and briefly compare it with alternative approaches such as the Holm and Benjamini–Hochberg (FDR) procedures. We also discuss the potential of dimensionality reduction techniques, such as principal component analysis, to decrease the number of comparisons. Furthermore, the article emphasizes the importance of research ethics, including transparency in reporting exploratory versus confirmatory analyses and avoiding questionable practices such as p-hacking. Finally, by reflecting on the role of academic presentations as preliminary reports, we suggest practical ways for researchers to enhance the statistical rigor of their studies. By raising awareness of these issues, this article aims to provide researchers—particularly those in applied and interdisciplinary domains—with a clear pedagogical foundation for avoiding common statistical pitfalls and for contributing to more reliable and reproducible scientific publications.
View full abstract