Exponential regression with correlated observations

Abstract

SUMMARY Finney's(1958) model for correlated observations y x in the exponential regression situation, E(y x )=α= β p x , is studied in greater detail. It is shown that his maximum-likehood procedure for estimating p is valid in only certain special cases. Some of his results for linear and quadratic estimators of p have been extended and his conjecture on the constancy of the asymptotic variance of members of a class of quadratic estimators is shown to be true.