10If covariances are known, Becker, Wu (2007) suggest an MRA that pools estimates of all regression parameters, not just of the focus variable, and that can be estimated with feasible GLS.