DECODING THE PUZZLING WORLD OF P-VALUES: A GUIDE TO NULL HYPOTHESIS TESTS AND RECOMMENDATIONS

Abstract

The conventional application of P-value hypothesis testing has long been a topic of concern within the scientific community, prompting the American Statistical Association (ASA) to proactively address these issues. In a landmark 2016 publication by Wasserstein and Lazar, the ASA outlined six foundational principles aiming to redefine the definitions and applications of P-values in hypothesis testing. This abstract explores the profound impact of these principles on statistical inference, scientific decision-making, and the interpretation of research results. The principles elucidate that P-values serve to indicate the extent of data incompatibility with a specified statistical model, explicitly disassociating them from measures of the probability of a hypothesis being true or the data arising solely from chance. Significantly, the principles emphasize that scientific decisions should not hinge solely on specific P-value thresholds, placing paramount importance on complete reporting and transparency in statistical inference. Moreover, the principles underscore that P-values do not inherently measure effect size or signify the significance of results, nor do they independently provide substantial evidence for a model or hypothesis. This transformative information presented by the ASA's principles holds the promise of resolving longstanding debates surrounding the utility of P-value hypothesis tests. Their adoption is poised to rectify misconceptions in textbook explanations, classroom instruction, and scientific paper interpretations related to hypothesis testing. The application of these principles has the potential to eliminate flawed thinking and language associated with P-value null hypothesis tests, assuming their continued prevalence in research and practice