Bounds for the bias of the LS estimator of o2 in the case of a first-order autorregressive process when the regression contains a constant term
Heinz Neudecker, Neudecker, H · AgEcon Search (University of Minnesota, USA) · 1976
The LS-estimate Ti7T of a 2 and its bias are considered, in the case of regression with a constant term. Following the procedure described in an earlier paper (AE2/76) 1 e'e much more satisfactory bounds are established for ---E a2 n-k. The computations show that the LS-estimator is biased toward zero for an impressive number of values of n, k and p.