A0251
Title: Robust optimal estimation of asymptotic covariance matrices in non-stationary multi-dimensional time series
Authors: Kin Wai Chan - Harvard University (United States) [presenting]
Abstract: Robust optimal estimation is considered for the asymptotic covariance matrix (ACM) of the sample mean in multi-dimensional time series that potentially have hidden trends and structural breaks. Robust estimation of the ACM is crucial to statistical inference because the ACM is naturally involved in many statistical procedures, e.g., change point detection, and construction of simultaneous confidence bands of trends. We propose an estimator of the ACM, which is robust to unknown forms of trends and possibly divergent number of change points. We also propose a robust estimator of the optimal bandwidth in a close form so that the estimator can be easily operated with an asymptotically correct optimal bandwidth. The estimator is robust, statistically efficient, computationally fast, and easy to implement. An empirical study on the S\&P 500 Index is presented.
