Multiple Comparisons | Understanding Medical Research

From the creator

Multiple comparisons are how a study of a drug that does nothing can still report a statistically significant result. This video covers why every study needs one pre-specified primary outcome, the Bonferroni correction, and why food frequency questionnaire studies deserve skepticism. Chapters: 00:00 Introduction 00:25 A fortune teller's trick 03:04 The 0.05 threshold and one hypothesis 04:17 Testing 50 hypotheses on a sugar pill 05:44 One primary hypothesis per study 06:43 Acupuncture for infant colic 08:43 The Bonferroni correction 10:43 ISIS-2 and astrological signs 13:20 The food frequency questionnaire 15:28 Take-home points Learning objectives: After watching, you'll be able to: - Explain why testing many hypotheses at a 0.05 threshold guarantees some false positives - Apply the Bonferroni correction to set a stricter significance threshold for multiple tests - Assess a study by locating its pre-specified primary outcome and how secondary results are framed 📺 Watch the full playlist on YouTube: https://www.youtube.com/playlist?list=PLh9mgdi4rNewVyTec_MVGxIbNSEeMhEsZ 🎓 Enroll on Coursera for assessments and a certificate: https://www.coursera.org/learn/medical-research 🌐 Explore more learning opportunities on Yale Online: https://online.yale.edu/courses/understanding-medical-research-your-facebook-friend-wrong

Published 2020-07-31 · Watch on YouTube ↗
Availability and playback restrictions are controlled by the original channel.