Abstract While a variety of new methods of monetary unit sampling have been developed over the last few years, steps have not been taken toward developing accompanying logic for selecting sample sizes. Without such procedures it is very difficult to take advantage of the increased efficiency that the new methods provide. This paper develops two sample size selection procedures for the Moment Method of Dworin and Grimlund: one approach focuses on simple types of sample results (e.g., zero or one error in the sample) and the other approach focuses on more general estimates of the total error in the population. Both approaches could be applied to other new methods. Besides presenting the two approaches, sample size tables, test results, and power curve-oriented simulations are provided. In order to demonstrate the relative efficiency of the two approaches, the sample sizes and power curves are compared to those of two forms of the Stringer bound.
Richard A. Grimlund (Tue,) studied this question.