This paper contrasts two model-based variance estimators for the simple regression estimator recently proposed in the literature. Both may be good estimators of the model variance as well as design consistent estimators of the design mean squared error. A third variance estimator that combines the strengths of the original two is also discussed.
