Change Point Detection via Self-normalization: A Personal Journey

Joint Event with the Department of Statistics and Data Science, College of Arts and Sciences

Abstract: Change point detection is a classical topic in Statistics with numerous applications. Motivated by the presence of structural breaks in data of expanding dimensions from fields like genomics, finance, and neuroscience, there has been a surge of interest in detecting change points within high-dimensional data characterized by intricate cross-sectional and temporal dependencies.

In this presentation, we will introduce a new nonparametric approach for testing and estimating change points in time series data using self-normalization. The discussion will begin with a review of self-normalization's role in inferring time series, focusing on its application in constructing confidence intervals and testing for changes in mean. Subsequently, we will briefly discuss recent publications concerning the segmentation of Covid-19 time series and high-dimensional change-point detection. Finally, our focus will be on dimension-agnostic change point testing. The proposed test is versatile across both low and high-dimensional settings, accommodating both temporal and cross-sectional dependencies.

If time permits, I will also touch upon ongoing work related to object-valued time series.

Bio: Xiaofeng Shao received his PhD degree in Statistics from the University of Chicago in 2006 and has since been a faculty member with the Department of Statistics at the University of Illinois Urbana-Champaign. His current research interests include time series analysis, change-point analysis, functional data analysis, high dimensional data analysis and their applications. He is a fellow of Institute of Mathematical Statistics (IMS) and American Statistical Association (ASA). He currently serves as an associate editor for Journal of Royal Statistical Society, Series B, Journal of the American Statistical Association and Journal of Time Series Analysis.

Host: Todd Kuffner