Econometrica: Jan 1971, Volume 39, Issue 1
Generalized Least Squares with an Estimated Variance Covariance Matrix
G. S. MaddalaThe paper discusses why certain commonly used two-step procedures give estimators which are asymptotically less efficient than the maximum likelihood estimator when there are lagged dependent variables among the regressors. This sort of problem is often encountered in the estimation of distributed lag models with serial correlation in the residuals.
Log In To View Full Content