A/B Testing pour ML

Peeking Problem

Bias introduced by repeatedly analyzing the results of an A/B test before its planned completion, which increases the risk of false discoveries (false positives) due to the inflation of the type I error rate.

← Quay lại