Once more, the whiskey has begun to flow. In this episode, Graham and Ray tackle the question of how many published studies are likely invalid or not reproducible. We learn about Type I and Type II statistical errors, publication bias, the Proteus Phenomenon and the Winner’s Curse. As well, the wrath of Almighty Zod rains down in the form of thunder, eliciting a visit from Graham’s adorable dog Ouie. And we learn that Ray’s Wikipedia page has been taken down. (Hey, he never created it in the first place, so nothing lost, right?)
Links from today’s show:
-
- Why Selective Publication of Statistically Significant Results Can Be Effective
- Why Most Published Research Findings Are False
- When science isn’t science-based: In class with Dr. John Ioannidis
- Scientific Regress
- The Cold Fusion claims of Stanley Pons and Martin Fleischmann
- The Complete Works Of Chuang Tzu (chapter 2)
Science Monkey 