Beyond Statistical Significance: Moving Past p< 0.05 : Advancing Inference with Second-Generation P-Values

Abstract

For decades, the p-value has been the dominant tool for statistical inference in biomedical and epidemiological research. Despite its ubiquity, the p-value has been widely criticized for fostering dichotomous thinking, encouraging misinterpretation, and obscuring clinical relevance. Second-generation p-values (SGPVs) have emerged as an alternative, offering a more nuanced approach by explicitly considering the overlap between confidence intervals and clinically defined thresholds of trivial effects. Unlike traditional p-values, which test a point null hypothesis, SGPVs incorporate an interval null, thereby distinguishing between inconclusive findings and true evidence of clinically unimportant effects. This paper outlines the limitations of conventional p-values, explains the framework of SGPVs, and discusses their advantages, challenges, and potential applications in clinical research, reproducibility studies, and high-dimensional data analysis. By shifting emphasis from statistical to practical significance, SGPVs promise greater transparency, interpretability, and rigor in scientific inference.