03/13/2023, 4:59 AM
Bayesian methods offer some distinct advantages over the frequentist approach. In order for frequentist methods to be valid, you must decide on the sample size in advance before running an experiment---a fixed horizon. (from the docs)
Bayesian vs frequentist , I know, I know) Just commenting one thing - calculating minimum nobs in frequentist approach let you get the exact numbers for Type 1 error and for Type 2 error, (sort of) exact numbers for minimum detectable effect and therefore clear algorithm for making decision in order to avoid people asking "so when should I decide which variation has won?".


03/13/2023, 7:32 AM
Yes, true - we could improve that language