A0524
Title: Robust reduced rank estimation for low-rank vector AR models
Authors: Fumiya Akashi - University of Tokyo (Japan) [presenting]
Abstract: The estimation problem of vector autoregressive models with possibly infinite variance error processes is considered. A general low-rank structure for the coefficient matrices is assumed, and an estimation procedure based on the multivariate median is proposed. The reduced rank estimation algorithm is also provided, and the proposed estimator is shown to improve the efficiency of the classical least-distance estimator in some sense. Some simulation experiments illustrate the finite sample performance of the proposed method.
