A0237
Title: Sequential Gaussian approximation for nonstationary time series in high dimensions
Authors: Fabian Mies - Delft University of Technology (Netherlands) [presenting]
Ansgar Steland - RWTH Aachen (Germany)
Abstract: To enable sequential inference in high-dimensional vector time series, Gaussian couplings of partial sum processes are constructed, for the regime $d=o(n^\frac{1}{3})$. The coupling is derived for sums of independent random vectors and subsequently extended to nonstationary time series. The new inequalities depend explicitly on the dimension and on a measure of nonstationarity and are thus also applicable to arrays of random vectors. A feasible bootstrap approximation scheme is proposed. To demonstrate the usefulness of the approximation results, applications to sequential testing and change-point detection are described.
