S (1) Suppose the model is: If the variance of the error term vector is: (2) Now suppose the model has first order serial correlation only: how would you estimate this model so that the estimator is the most efficient? 4. Heteroskedasticity and his twin brother Serial Correlation have come back! Var (& |X) = E (EET (X ) = V = diag {o? } = ticity and serial correlation, please derive the Generalized Least Squares estimator, the B Yt = XTBo + Et, 04 1- 43d - 73 ' 93 + 9 3 X = 3 estimator? Particularly, what's the difference between Prais-Winsten and Cochrane-Orcutt? Yt = 68 + 60 Xit + 62X2t + Et, t= 1, . . . ,n. We assume |p|