Annals On Call Podcast · American College of Physicians

Improving Research Reports: Avoiding p-Values

April 1, 2024·24 min·2 clips

Dr. Bob Centor discusses how misclassification impacts statistical evaluations like number needed to treat.

This episode of Annals On Call features host Dr. Bob Centor and guest Dr. David Savitz, a professor of epidemiology at Brown University, discussing a published opinion piece on statistical significance testing. They focus on the challenges researchers face when responding to editorial feedback about statistical methods. The conversation centers on moving beyond traditional reliance on p-values in medical research reporting. The article discussed is titled "Responding to Reviewers and Editors about Statistical Significance Testing." Dr. Savitz argues that p-values and statistical significance testing are often misunderstood and misused. He advocates for researchers to emphasize estimation, such as confidence intervals, over binary significance testing. The discussion covers how to communicate with editors and reviewers who insist on traditional p-value thresholds. Savitz explains the limitations of p-values in conveying the magnitude or precision of an effect. He addresses common scenarios where dichotomous thinking based on p-values can be misleading. A key insight is that a p-value above 0.05 does not mean a study found "no effect," but rather indicates less precise evidence. Savitz suggests framing results around the range of plausible effects shown by confidence intervals. He notes that the American Statistical Association has released statements cautioning against over-reliance on statistical significance. The conversation highlights how focusing on estimation can improve the interpretation and clinical applicability of research findings. Savitz provides practical advice for authors on how to respectfully push back against requests for inappropriate statistical re-analysis. The tone is educational and conversational, with the host and guest exploring technical concepts in an accessible manner. The style is advisory, offering guidance to clinician-researchers on navigating publication standards. This episode would be highly valuable for academic clinicians, researchers, and students who prepare manuscripts for medical journals. Listeners with no interest in research methodology or statistical debate might find the content too specialized.

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Annals On Call Podcast