Stefan Faridani
@sfaridani.bsky.social
Assistant Professor of Economics at Georgia Tech / Econometrics, Meta Analysis, Spillovers
I assign readings from this book and my students not only actually do them (!) but also tell me that the readings really help them. Awesome text!
June 28, 2025 at 9:47 PM
I assign readings from this book and my students not only actually do them (!) but also tell me that the readings really help them. Awesome text!
Reposted by Stefan Faridani
That, or something similar, under a formal NDA, is what I do several times a year.
May 19, 2025 at 11:19 PM
That, or something similar, under a formal NDA, is what I do several times a year.
Under the counterfactual being studied we would also double the number of treatment clusters! So it really appears that lots of RCTs are either overpowered or are chasing something close to a true null. Exactly 80% powered studies seem quite rare. Here is the paper if you are interested:
January 27, 2025 at 1:33 PM
Under the counterfactual being studied we would also double the number of treatment clusters! So it really appears that lots of RCTs are either overpowered or are chasing something close to a true null. Exactly 80% powered studies seem quite rare. Here is the paper if you are interested:
My JMP was on exactly this subject! I estimated that doubling the sample size of every econ rct would only increase power by 7 pp on average. It turns out that in econ the true power of most RCTs (conditional on the true effect) is either very high (little benefit to doubling) or very low.
January 27, 2025 at 12:49 AM
My JMP was on exactly this subject! I estimated that doubling the sample size of every econ rct would only increase power by 7 pp on average. It turns out that in econ the true power of most RCTs (conditional on the true effect) is either very high (little benefit to doubling) or very low.