Bounds for the bias of the LS estimator of o^2 in the case of a first-order autoregressive process

The LS-estimator Ti7.7i of a 2 and its bias are considered. Bounds are then established for 2- by means of a well-known theorem by Anderson. a2 n-k The same bounds can be derived by a direct applicant of the Courant-Fischer min-max theorem. Following Anderson and Grenander-Szeg6 one can find approximations to these bounds.