Cracking Correlated Observations In A/B Tests With Mixed Effect Models
A/B testing is the workhorse for many of our product decisions. A/B testing seemed easy to me at first glance: split up observations, compute a metric, then compare! But whether it’s problems with early…