Conditional uncertainty: Misinterpretations of “significant” p values

Abstract

Misapplication of p values can lead to very misleading conclusions. The frequent insignificance of a “significant” p value 1provides examples illustrating situations where p values are misused in interpretation and describes alternatives to p values to consider. The concern about p values, and their potential shortcomings and misinterpretations, has been widely discussed and has received a lot of debate from the statistical community. While we agree with much of the content presented in the article, the issue fundamentally represents a misinterpretation of conditional probability, misapplication of measures regarding diagnostic accuracy, and consequently violations of assumptions inherent in experimental design and the validity of many approaches to statistical analysis and inference. One way this can be remedied is through team science, and collaboration between clinical and biostatistical scientists in research.