Asymptotic efficiency of the OLSE for polynomial regression models with spatially correlated errors

Abstract

For polynomial regression models with spatially correlated errors, the covariance matrix of the ordinary least-squares estimator (OLSE) is shown to have the same limiting value as that of the generalized least-squares estimator (GLSE) under the same normalization. This implies that the OLSE is asymptotically efficient. © 2000 Elsevier Science B.V.