Systematic Solutions to the AA Problem in Random Experiments

Speaker

Wanbo Kui – Didi Data Analyst

Graduated in June 2021 with a B.Sc. in Statistics and Data Science from Southern University of Science and Technology; earned a M.Sc. in Statistics and Data Science from National University of Singapore in January 2023; has been working at Didi’s data science platform since January 2023, focusing on optimizing all stages of random split experiments.

Talk Title

Systematic Solutions to the AA Problem in Random Experiments

Talk Introduction

Although A/B testing is the gold standard for decision making, its validity is compromised when the AA problem arises. Various solutions exist, and combining re‑randomization with regression adjustment is among the most effective, helping to mitigate the AA issue, prevent it proactively, and increase the credibility of experimental results.

Talk Outline

1. Academic and industrial research on re‑randomization

2. Demonstration of re‑randomization principles and data simulation

3. Practical applications of re‑randomization and key considerations

Audience Benefits

1. Understand recent advances in covariate balance

2. Familiarize with the underlying mechanisms of re‑randomization

3. Master how to apply re‑randomization in practice to alleviate the AA problem

Click "Read Original" to view offline conference details~