An Extension of the Moment Method in Statistical Estimation

摘要

In parameter estimation problems, a distinct advantage associated with the method of moments over the maximum likelihood method is its practical expediency. That the method has enjoyed little popularity stems mainly from the fact that it generally leads to inferior estimates in terms of low asymptotic efficiency. In this paper an optimal scheme of combining a set of moment estimators is considered. It is shown that the scheme leads to an estimator offering computational simplicity as well as increase in efficiency under general conditions. (Author)