【解説】多重検定の問題とは | 仮説生成と仮説検証

The theme of this sixth episode is "The Problem of Multiple Testing." The P-value is often the focus of attention when performing statistical hypothesis testing. However, a low P-value does not necessarily mean that the alternative hypothesis is true. "Multiple testing" can produce low P-values. In this video, we explain why multiple testing can be problematic and in what situations it should be considered problematic. We've included a review of the previous episode at the beginning of this video, but for more information on inferential statistics and hypothesis testing, we recommend that you refer to the previous video (   • 【解説】統計学的に有意とは｜P値と仮説検定  ) and the video before that (   • 【重要】推測統計とは何か？～標本と母集団～｜論文結果の「過度な一般化」に注意  ). My voice gets a little hoarse towards the end, but that's just how it is... ■Table of Contents 00:24 This Time's Conclusion 00:50 What is the Problem with Multiple Testing? 01:11 Review of Previous Lessons | Inferential Statistics 01:49 4 Steps to Hypothesis Testing 02:18 What Does "Sufficiently Low P-Value" Mean? 03:00 Alpha Error 03:24 Example of Multiple Testing 05:00 Correction for Multiple Testing | Bonferro's Method 05:37 Various Methods for Correcting Multiple Testing 06:11 Summary 1 06:46 Should Alpha Error Be Within 5%? 07:08 Hypothesis Testing 07:59 Hypothesis Generation 08:50 Examples of Hypothesis Testing and Hypothesis Generation 09:38 Examples from Actual Papers 11:43 This Time's Conclusion 12:28 Bonus | Non-Independent Hypotheses ■Skimming "Medical Statistics in the Skimming" is a channel that explains medical statistics in a short, easy-to-understand manner for clinicians, residents, medical students, nurses, pharmacists, rehabilitation staff, etc. ■Subscribe to the channel here    / @sukimaru-stats   ■Channel Creators This channel is run by two internists working at a community hospital. Blog and Twitter are here. ◇Blog 🧠Nounai Librarian https://nounai-librarian.com 🍤Ebikatsu Yokocho https://evineko.com/ ◇Twitter 👴Statistics Uncle   / medibook3   👵Evidence Auntie   / evidenceobasan