Title: Estimation of the long-run error variance in nonparametric regression with time series errors
Authors: Marina Khismatullina - University of Bonn (Germany) [presenting]
Michael Vogt - University of Bonn (Germany)
Abstract: A new difference-based estimator of the long-run error variance for nonparametric regression is proposed in the case that the error terms have an autoregressive structure. Such an estimator is required for virtually all inferential procedures in the context of nonparametric regression. Our proposed estimator improves on existing methods in several respects. First, the estimator produces accurate estimation results even when the AR process is quite persistent. Second, it produces accurate results even in the presence of a very pronounced regression function. These properties are illustrated by a simulation study that compares the proposed estimator with existing ones.