A1555
Title: Forecasting under breaks in non-parametric regression
Authors: Sze Him Isaac Leung - The Chinese University of Hong Kong (Hong Kong) [presenting]
Chun Yip Yau - Chinese University of Hong Kong (Hong Kong)
Abstract: The prediction of future observations in a non-parametric regression model which is subject to a structural break in time is studied. We propose a weighted kernel estimator to estimate the post-break function, where the weights are both time and location dependent. It is shown that incorporating pre-break observations can improve the estimate of the post-break function in terms of the mean squared forecast error (MSFE). This is related to the bias-variance trade-off induced by including pre-break observations. Simulation studies indicate that the proposed weighted kernel estimator has a lower MSFE compared with traditional post-break methods.
