search

aaronpeikert / bayes-prereg

https://aaronpeikert.github.io/bayes-prereg/#2
Creative Commons Zero v1.0 Universal
6 stars 6 forks source link

control issues #95

Open aaronpeikert opened 10 months ago

aaronpeikert commented 10 months ago

Furthermore, once again, we do not 'increase' error rates, we control them.

Indeed, based on the example on page 13, it seems the authors themselves confuse real theoretical risk with Type 1 error control.

We do not want to reduce the error rate inflation in preregistration - we want to control it.

If the error rate is not controlled, I do not know how severely a claim has been tested, and I can not use the claim to decide how to act as a scientist. The authors seem happy with the fact that preregistration reduces the Type 1 error rate even in an exploratory study. But I am not.