News for hackers. Stuff that matters.
Re:I was deceived about Covid vaccine safety by dekhn on Monday January 24, 2022 @06:16PM (#30065496)
yeah, I don't think that's a useful insight. I would say that if a result is not statistically significant, you shouldn't consider it to be the "best available estimate" (which is sort of a different concept anyway) unless you're willing to deal with a very high false positive rate.

BTW I set my p threshold at 1e-6 for a single test, 0.05 is just crazy. On the other hand, I've worked with systems where nobody cared about the significance, just that the weights that got pushed to production made more money than the last set of weights.

There are no comments on this comment yet.