Tag: statisticalpower

A/B testing, zeroinflated (truncated) distributions and power
Naive A/B testing just uses ttests or proportion tests, with the assumption that at large sample sizes, the right statistical test does not matter that much. I explore the case of a zeroinflated upperbounded Poisson distribution and find that using the wrong test can require 3x the sample size to achieve the same statistical power, a difference large enough to matter in a real business setting.