Shenzhen University School of Mathematical Sciences
Liyuan Scholars Colloquium Session 188
Title: A General Stabilization Framework to FDR Control with Applications
Speaker: Professor Zhong Wei (Xiamen University)
Time: 15:30–16:30, November 22, 2026
Location: Room 601, Huiyan Building, Yuehai Campus, Shenzhen University
Abstract: Stability is a basic requirement for reproducible data analysis: conclusions should not be overly sensitive to small changes in the data or the way they are analyzed. This talk focuses on stable discovery under false discovery rate (FDR) control. FDR control limits the expected proportion of false discoveries, but selected sets can still vary across repeated analyses of the same data due to randomness of the algorithms. We present the FDR Stabilizer, a general framework for improving stability in multiple testing and variable selection. The method combines evidence from repeated analyses into a single ranking and determines the selection size from the rounded average number of discoveries. Finite-sample bounds compare its FDR and power with those of the original procedure, while asymptotic results establish conditions for FDR control without power loss. We also study how the selected set stabilizes as the number of repetitions increases. Two extensions illustrate further developments of this framework. In skeleton learning for high-dimensional directed acyclic graphs, false edges can mislead network interpretation and subsequent causal analysis, while unknown conditioning sets and dependent tests complicate FDR control. In multiple change-point detection, false alarms can suggest structural changes that never occurred. Candidate locations also vary across analyses, requiring stabilization of both the number and locations of detected changes. We develop extensions for both problems, with more stable selections and asymptotic FDR guarantees.
Speaker Profile: Zhong Wei is a Nanqiang Distinguished Professor (Level 2) and doctoral advisor in the Department of Statistics and Data Science at the Wang Yanan Institute of Economics and the School of Economics, Xiamen University. He is a Leading Talent in the National Major Talent Program (Ministry of Education, 2024) and a recipient of the National Natural Science Foundation of China’s Outstanding Young Scientist Award (2019). He also serves concurrently as Director of the Office of International Cooperation and Exchange, Director of the Office of Taiwan, Hong Kong, and Macao Affairs, and Director of the Office of Overseas Education Programs at Xiamen University. His research focuses primarily on statistical analysis of high-dimensional data, statistical learning algorithms, and econometrics. He has served as an Associate Editor (AE) for six domestic and international journals, including JASA and ISR, and has published (or had accepted) more than 50 papers in leading domestic and international statistical journals such as AOS, JASA, Biometrika, JoE, JBES, JMLR, AOAS, Science in China—Mathematics, and Statistical Research. He has received the Fok Ying Tung Education Foundation Young Scientist Award for Higher Education Institutions, the 9th Outstanding Achievement Award for Scientific Research in Higher Education Institutions, the National Baosteel Outstanding Teacher Award, the Grand Prize in the Xiamen University Young Faculty Skills Competition, was named one of Xiamen University’s “Ten Most Beloved Teachers,” and serves as the leader of the Nanqiang Excellence in Teaching Team at Xiamen University. He is the chief editor of Statistical Foundations of Artificial Intelligence, a textbook published under the Ministry of Education’s “Statistics 101” Program.
All faculty and students are welcome!
School of Mathematical Sciences
November 9, 2026