Further remarks on exponential regression with correlated observations

Abstract

Mitchell (1968) studied the properties of estimators of p in the exponential regression situation, E(y x )= α+βp x , when the observations are assumed correlated. In the notation of that paper it is now shown that the quadratic estimators r(O, c, d) have uniform minimum asymptotic variance among all linear and quadratic estimators of p and that estimators can be excluded from this class on grounds of possessing uniformly worse asymptotic biases.