|
Prior Information on the Coefficients when the Disturbance Covariance Matrix is Unknown
William E. Taylor
Abstract
In a linear regression model with arbitrary disturbance covariance structure, least squares estimators subject to correct linear restrictions dominate unrestricted least squares for all estimable functions of the parameters if and only if the covariance matrix obeys conditions closely related to those of the Gauss-Markov theorem.
|
