Common misinterpretations of statistical significance and P-values in dairy research

Abstract

Careful communication of results is integral to dairy research. However, many published studies contain misinterpretation of the results of statistical analysis, which can lead to conclusions being drawn which are not consistent with the data. Many of these interpretations have arisen because of a focus on P- values rather than on the potential range of effects that are compatible with the study data. This review focuses on 3 misinterpretations: the use of levels of statistical significance to compare results between or within studies, overinterpretation of nonsignificant results, and the use of “trend” to describe results that are “close” to a significance threshold. All of these misinterpretations can be avoided by paying more attention to the range of effects that are compatible with the data. Such a focus will have many benefits—not least, making it clearer when studies have insufficient power to accurately characterize their outcomes. Focusing on compatible effects is not a panacea but will improve statistical inference and provide more thoughtful descriptions of study outcomes.