A1304
Title: Fully functional sieve covariance inference of locally stationary functional time series
Authors: Yan Cui - Reed College (United States) [presenting]
Abstract: Simultaneous statistical inference is established for the auto-covariance functions of locally stationary functional time series based on full functional information rather than employing dimension reduction techniques. The sieve method is leveraged to estimate the unknown auto-covariance function with flexible choices of orthonormal basis functions. A fully functional multiplier bootstrap methodology is proposed to construct asymptotically correct simultaneous confidence regions for the auto-covariance functions, which can be validated by a uniform Gaussian approximation over all Euclidean convex sets for sums of a class of moderately high-dimensional locally stationary time series. The proposed approach is applied to an air quality functional time series dataset to investigate the variability of the auto-covariance functions.
