Date: Friday, October 3, 2025 - 12:00pm to 1:30pm ET
On Friday, October 3, 2025, the National Institute of Statistical Sciences (NISS) hosted a timely and engaging installment of the AI, Statistics and Data Science in Practice Webinar Series, titled Covariate Adjustment, Intro to Resampling, and Surprises. The virtual event, held from 12:00 to 1:30 pm ET, brought together statisticians, data scientists, and practitioners to examine practical methods for improving the accuracy, efficiency, and interpretability of experimental results, particularly in the context of A/B testing.
Covariate Adjustment in Practice
The webinar was led by Tim Hesterberg, Staff Data Scientist at Instacart, who offered a clear and accessible discussion of covariate adjustment as a relatively simple yet powerful tool for increasing experimental precision and reducing bias. While randomized experiments are unbiased under formal statistical definitions, Dr. Hesterberg challenged participants to consider bias from a more intuitive, real-world perspective, highlighting why adjustments can still meaningfully improve results in practice.
Resampling Methods and Statistical Surprises
A central focus of the talk was resampling methodology, including bootstrap and jackknife techniques, as tools for assessing uncertainty and measuring accuracy when analytical approaches fall short. Dr. Hesterberg illustrated how these methods can reveal unexpected behavior in common assumptions, including the limitations of traditional normal approximations and the often-cited n ≥ 30 rule. Through concrete examples, he demonstrated how resampling can expose inaccuracies that might otherwise go unnoticed, offering both cautionary lessons and practical guidance for applied data analysis.
Moderation and Audience Engagement
The session was moderated by Yijun (Frank) Wei, Senior AI/ML Scientist at General Motors, who facilitated discussion and guided audience questions. The webinar prompted active engagement from attendees and underscored the continuing importance of sound statistical foundations in modern data science workflows, particularly as experimentation and algorithmic decision-making scale across industries.
Thank You & Acknowledgements
NISS extends its sincere thanks to Dr. Tim Hesterberg for sharing his expertise and practical insights, and to Yijun (Frank) Wei for his thoughtful moderation of the session. We also gratefully acknowledge all webinar participants for their engagement and contributions to the discussion. Finally, NISS thanks the AI, Statistics and Data Science in Practice Webinar Series organizers and staff for their continued efforts in delivering high-quality programming that bridges statistical theory and real-world application.