A Portmanteau White-Noise Test for Multivariate Locally Stationary Functional Time Series
A portmanteau-type hypothesis test of the white-noise (serial-uncorrelatedness) assumption for multivariate locally stationary functional time series, working directly on the curve-valued data without dimension reduction and aggregating over an increasing number of lags. Because the null limiting distribution can be non-standard or nonexistent, calibration uses a bootstrap. The theory rests on a new Gaussian approximation for the maximum of degenerate second-order U-statistics of functional time series.
2501.00118
Lujia Bai, Holger Dette (Ruhr-Universitaet Bochum) and Weichi Wu (Tsinghua University) introduce a portmanteau-type white-noise (serial-uncorrelatedness) test for multivariate locally stationary func…