In the session, we will explore the enduring relevance of traditional regression methods in modern experimentation, focusing on regression adjustments, CUPED and their variations. Attendees will gain a deep understanding of how these techniques leverage pre-experiment data to reduce variance and increase the sensitivity of experiments, even in the era of advanced machine learning and causal inference methods. We will discuss the theoretical foundations of regression adjustments, practical applications of CUPED, and how these methods compare to newer approaches in terms of robustness, interpretability, and scalability. Key takeaways include learning how to implement these techniques effectively, understanding their limitations.
Technical level: Introductory level/students (some technical knowledge needed)
Session Length: 40 minutes
Supported by
