Hi! I wonder what method you use to estimate the relative uplift in confidence intervals in Frequentist engine? My understanding is that it’s not possible to do that with parametric method since the distribution of the uplift doesn’t have a defined variance and expected value (Cauchy distribution)?

There are many common approaches to estimating relative change in frequentist testing. We use the delta-method derived estimate of the variance to build our confidence intervals.

Thanks! I should probably familiarize myself with it. Didn’t notice this mentioned in the docs Graham provided, but indeed it’s there.

It's not very visible. I'd like to write more detailed statistics docs, but unfortunately we're prioritizing other features at the moment 🙂