2025-03-07 11:30:03
Re @pat_browne11 Statistics good! I was reacting to the attitude many people take that formal statistical significance is the *only* measure of truth. In practice they're (usually) imperfect approximations. Useful, but just one view, to be combined with data vizualization, intution, etc.
2025-03-07 06:41:36
Re @domenic @LuisCostigan1 @LuisCostigan1 my complements on this (and http://n1.tools in general)! This is makes me a bit more optimistic. I'm actually tempted to try this again with coffee, since coffee in particular makes me nervous (unlike tea) and I rarely drank it during this experiment.
2025-03-07 06:36:15
Re @domenic @LuisCostigan1 Interesting! I think this is it? I'm surprised the posterior was so confident with such limited data (83%). But both the protocol and the analysis/model look good to me...
https://manifold.markets/LuisCostigan/nof1-blinded-experiment-will-210mg
2025-03-07 03:21:29
Re @onemanstartup Mostly because I didn't think of it! (And slightly because I don't have room in my life for a smartwatch.) I wish I had HRV data, though I'd probably also record subjective stress in any case.
2025-03-07 02:13:07
Re @BigMandelbrot Well except if you REALLY believe all the p-values, then you'd have to accept that end stress was *higher* after theanine (p=0.029) and that D felt "more like theanine" than theanine (p=0.029)!
(The p-values are slightly different if you look at more digits.)
