Shrinkage Estimation in Seemingly Unrelated Regression Models
Friday, April 12, 2019
Continuing Education Building - Room 2030 (University of California, Irvine)
*Names in bold indicate Presenter
One of the issues in the seemingly unrelated regression models is whether to assume heterogeneity or homogeneity of the parameters of interest. In this paper we propose a shrinkage estimator which is a Stein-like estimator and is the weighted average of a restricted and unrestricted estimator. Similar to Hansen (2015), the weight depends on a statistic that measures the distance between these two estimators. We study the bias and mean squared error of the proposed estimator and compare them with those of the unrestricted, restricted and other estimators proposed in the literature using both the small-disturbance theory and asymptotic theory. The conditions for dominance over the unrestricted estimator under the mean squared error are obtained.