Second-order risk comparison of SLSE with GLSE and MLE in a regression with serial correlation (Q1062705)

First, with a series of lemmas, propositions and theorems the author gives the second order bias of the estimate of the autoregressive parameter based on the ordinary least squares residuals in a linear model with serial correlation. Next, with additional lemmas and propositions, he gives the second order expansion of the risk matrix of the generalised least squares estimate (GLSE) based on the estimated residuals by the simple least square estimate. Further, he shows that the risk matrix of the GLSE is asymptotically equivalent to that of the maximum likelihood estimate up to the second order. Also a sufficient condition is given for the term, due to the estimation of the autoregressive parameter, in this expansion, to vanish under \textit{U. Grenander}'s condition [Ann. Math. Stat. 25, 252-272 (1954; Zbl 0056.382)]. Finally, some illustrative examples are given.