4 ms·
>I run an A/B test and see statistical significance That’s exactly what power gives you: a fighting chance to detect anything. Not sure where the misunderstand
by thelamest 5y ago
>I run an A/B test and see statistical significance
That’s exactly what power gives you: a fighting chance to detect anything. Not sure where the misunderstanding comes from – it’s about the worthwhileness of tests themselves, not “features”.
You may think about power visually by relating them to confidence intervals – higher power → more precise (expected) estimate. Low power → bands so wide that you can just as well use an RNG.