Single Versus Double-Sided Hypotheses and Probabilities

Abstract

Single-sided (1-tailed) and double-sided (2-tailed) probabilities are products of statistical tests that can be crucial to drawing accurate conclusions in scientific studies. In a review of articles published in issues of Pediatric Emergency Care from 2020, we identified 2 where single-sided versus double-sided probability issues potentially reversed a conclusion of study investigators. The purpose of this study is to describe single-sided versus double-sided probability issues found in Pediatric Emergency Care 2020 articles to increase awareness surrounding these issues.

Methods: This study involved a review of all articles from 2020 issues of the Pediatric Emergency Care journal, examining whether P values between and including the values 0.05 and 0.10, were characterized as not significant when, in fact, they resulted from a double-sided test and arguably should have been halved to yield significant single-sided probabilities less than or equal to 0.05.

Results: Two such studies were identified. In the first study, researchers concluded that their intervention resulted in "no statistically significant improvement," citing a P value of 0.08, but if a single-sided P value was used, it would have been 0.04 and the authors would have instead concluded that their intervention resulted in significant improvement. In the second study, researchers measured resuscitation times in pediatric and adult manikin simulations. They concluded no difference, citing a P value of 0.088, but if a single-sided P value was used, it would have been 0.044, and the authors would have instead concluded that the resuscitation times took longer in the pediatric simulation.

Conclusions: These articles demonstrate how single-sided versus double-sided probability issues can cause researchers to draw inaccurate conclusions. As such, we would urge that this be more rigorously evaluated when the P values are between 0.05 and 0.10.