Virtual Efficient Estimation of Treatment Effect in Online Experiments
Wed, Nov 30, 2022 @ 10:00 AM - 11:00 PM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Dr. Congshan Zhang, Meta , Core Data Science Team at Meta
Talk Title: Efficient Estimation of Treatment Effect in Online Experiments
Abstract: Randomized controlled trials are commonly used by tech companies to draw causal conclusions on various product changes. The confidence intervals from these experiments, however, are usually too large due to reasons such as limited number of users, heavy-tailed outcome variables and small treatment effects. Improving estimation efficiency for randomized controlled trials is not only a scientifically interesting but also a practically relevant area of research. In this talk, I will go over a few prominent techniques in statistics to improve estimation efficiency. Basic techniques such as CUPED and more advanced methodologies based on ML and synthetic controls will be introduced.
Biography: Congshan Zhang is a research scientist on Core Data Science Team at Meta. Congshan is interested in various topics in statistics and econometrics including causal inference, machine learning and time series. Congshan holds Ph.D. in economics from Duke University. Before joining Meta, Congshan did research on financial econometrics, with a focus on nonparametric and semi-parametric inference using high-frequency data and on testing models of financial markets. His work appears in top journals of econometrics such as Journal of Econometrics and Annals of Applied Probability.
Host: Urbashi Mitra
More Info: https://usc.zoom.us/j/96927080167?pwd=Vk9MOEpOSUx3V1hlZFc3U0tmOTNsUT09 Meeting ID: 969 2708 0167 Passcode: 586135
More Information: ECE Seminar Announcement_Nov21.docx
Location: https://usc.zoom.us/j/96927080167?pwd=Vk9MOEpOSUx3V1hlZFc3U0tmOTNsUT09 Meeting ID: 969 2708 0167 P
Audiences: Everyone Is Invited
Contact: Susan Wiedem
Event Link: https://usc.zoom.us/j/96927080167?pwd=Vk9MOEpOSUx3V1hlZFc3U0tmOTNsUT09 Meeting ID: 969 2708 0167 Passcode: 586135