Psychology · Research Methods & the Replication Crisis
A topic within Research Methods & the Replication Crisis, itself one of 10 topics in that field and part of Psychology.
2 works
John P. A. Ioannidis
Argues that even a genuinely true effect, once it clears a significance threshold in an underpowered or heavily tested literature, will on average have its estimated size inflated by the same selection process that produces false positives.
Notes on this copyKatherine S. Button, John P. A. Ioannidis, Claire Mokrysz, Brian A. Nosek, Jonathan Flint, Emma S. J. Robinson & Marcus R. Munafo
A study underpowered to detect the effect it is looking for will, on the rare occasions it does find something significant, systematically overestimate how large that effect is - so low statistical power inflates the effect sizes that make it into print.