Time:
15:30-17:00, September 1, 2020 (Tuesday)
Venue:
Tencent Conference ID: 578 184 484
Speaker:
Yao Fang, professor of Peking University, director of Peking University Statistical Science Center
Summary:
For analysis of spatial temporal data from a functional perspective, a heuristic extension of Karhunen-Loeve expansion is often used to decompose such data into temporal components and spatially correlated random fields. This structure provides a convenient tool to investigate the space-time interactions, but may not always hold for complex situations. In this work, we introduce a new concept of weak separability, and propose formal testing procedures to examine the validity of the heuristic Karhunen-Loeve decomposition. Asymptotic properties are studied to avoid using resampling procedures, e.g. bootstrap. Both parametric and nonparametric approaches are developed to estimate the asymptotic covariance by constructing lagged type estimators. We demonstrate the efficacy of our method via simulations, and illustrate the usefulness using two real examples: Harvard forest data and China PM2.5 data.
Brief introduction of the speaker:
Professor of Peking University, director of Peking University Statistical Science Center, and the fellow of Institute of Mathematical Statistics (IMS) and the American Statistical Association (ASA). He graduated from the University of Science and Technology of China with a bachelor's degree in statistics in 2000, and received a PhD in statistics from the University of California, Davis in 2003. He was a tenured professor in the Department of Statistics at the University of Toronto. He is currently the chief editor of the Canadian Journal of Statistics, and has served as an editorial board member of 9 core international statistical journals, including the top journals of statistics, Journal of the American Statistical Association and Annals of Statistics.
