eunho / sampledatasize Goto Github PK

10% relative detectable difference, 10% significance level, 80% Power:

15% relative detectable difference, 10% significance level, 80% Power:

Avoid PEEKING to stopping the test as soon as the results "look" significant Significance level(Type I error): Probability of the difference between control and challenger will be detected, when actually it does NOT exist. Statistical power(1-Type II error): Probability of the difference between control and challenger will be detected, when actually it exists.