Not at all. That’s me as a monster. 😂

👹

Technically, yes, but the comparison group of estimators is the same as before. We don’t gain anything by knowing it’s BUE in the class of linear, unbiased estimators.

In the end, it was an incorrect interpretation of what Hansen had shown. He proved OLS is BUE but failed to be careful about the class of estimators in the comparison group. OLS is still only BLUE because there are no nonlinear competitors when we restrict the class of estimators appropriately.

If one adds estimators that are unbiased under G-M, then there are nonlinear, unbiased estimators in play. And now OLS is no longer best. OLS does not need homosk and no ser corr for unbiasedness. It's not too surprising that estimators that use these assumps can be more effic.

Hi Jason. It's a bit subtle. If the comparison group only includes estimators that are unbiased under E(Y|X) = X*b and the rank condition on X, then any such estimator must be linear in Y. In a sense, we can say "OLS is BUE," but it's vacuous because there aren't any nonlinear, unbiased estimators.

😂😂

I almost met these criteria.

But I love regressions with lots of interactions!

How about you move to MSU and then your husband can manage Eagle View Golf Course? And bring a real pub to mid-Michigan. 😬