The headline is broadly defensible, but the qualifications matter. Effect sizes vary by population, the strongest claims rest on shorter trials, and credible voices push back on how it's typically framed.
they have uh an Achilles heel that is not there for let's say parallel groups where it's just some get a diet A and some get diet B and that is you can have what's called carryover effects and it turns out I've started to study this I I tip my hat to David Ludvig who pointed the issues out to me and I hadn't fully comprehended them before that uh there's almost no way around it right
Every Sunday: the week’s new conflicts and verdict changes — and nothing else.
Native comments, Twitter mentions, and Reddit threads about this claim — surfaced together so the conversation isn't fragmented across platforms.
Bookmarking — the dossier-vs-overview split is the right call. Most of the time I want overview; sometimes I want receipts.
Would love a "what would change this verdict" RSS feed. Sign me up if it exists.
So, I'm guessing that if Kevin I don't know, I don't follow this debate, by the way, but I'm going to guess that you're going to say Kevin Hall favors a crossover. And I would, not to speak for Kevin, but I would bet the reason Kevin favors it is because the type of work Kevin does is insanely expensive. He's putting patients in metabolic chambers and therefore the fewer patients that he needs to do that with, the easier he can do his work. Is that basically the argument? >> You're absolutely correct. And when you said statistics in the or power rather in the beta sense you're actually that is what it is. Uh it is statistical power the probability of rejecting the null hypothesis if the null hypothesis is false which is one minus beta the type two error rate. U there's actually a little bit more but I think 90 90 plus% of the motivation is what you've described. The other, as actually Kevin pointed out to me in some dialogue we were having was sometimes it's just throughput. Even if I had infinite money, I can't put people through the procedure fast enough because, you know, there's only so many chambers. >> Yep. >> Uh someone else pointed out to me it could be patient availability. If you said, I'm studying this in this rare population. >> I can't get a thousand people even if I had the money because they don't exist. I could only get 10 or 20 people. So all of those things strongly favor the crossover which in almost all circumstances will be much more statistically powerful. Meaning you can get the same amount of precision >> out of many fewer subjects.