Choosing a Stats Engine
This page has moved. The stats-engine comparison, the calculator modes, and the calculator's four measures now live in one place:
There used to be two versions of this page, and they had drifted apart. This one still claimed the sample-size calculator only worked for simple conversion metrics. That has not been true for some time: it now sizes conversion rates, rates, percentiles, and composites too. Rather than keep two copies in sync, there is now one.
Related
- Choosing a Stats Engine is the full comparison: the calculator, and when to pick each engine.
- The Frequentist engine is A vs B's classic method: the p-value (the odds of seeing this by chance alone), plus α and multiple-comparison corrections.
- The Sequential engine is built for safe early stopping: always-valid inference, checked as often as you like.
- The Bayesian engine is A vs B's default: probability to beat control, credible intervals, and the prior (what you expect before the test runs).
- Comparing engines: read the same data under all three engines, side by side.
- Analysis Defaults: set the project-wide engine default.