A0787
Title: Asynchronous change point inference in high-dimensional locally stationary time series
Authors: Cheuk Hin Cheng - The Chinese University of Hong Kong (Hong Kong) [presenting]
Kin Wai Chan - The Chinese University of Hong Kong (Hong Kong)
Abstract: Change point analysis is important for applications in fields such as biomedical, financial, and environmental systems, where high-dimensional time series often exhibit nonstationarity, with random shifts commonly occurring across components. Existing methodologies, restricted to stationary time series and synchronized change point assumptions, fail to address these challenges. A nonparametric Z-estimation framework is proposed, tailored to locally stationary high-dimensional processes that (i) tests structural break, (ii) estimates the mean and variance of the change point location, and (iii) infers break magnitudes across dimension. Theoretically justified and simulation validated, the approach robustly detects asynchronous shifts under complex temporal dependence. The application demonstrates its effectiveness in identifying localized disruptions, advancing robust inference for modern high-dimensional time series with heterogeneous change patterns.
