Researcher, Zalando SE
1 paper at NeurIPS 2025
We propose a new sequential test for online experiments that permits early stopping without increasing false detection risk and outperforms existing methods in statistical power, as shown in semi-synthetic simulations.