부산대학교 도서관

Sieve sampling is a list sequential, monetary-unit selection method which exploits the natural line item structure of the population and can easily be implemented either by hand or on a computer. It selects distinct line items and has been shown to estimate the total misstatement amount in substantive testing with good results. It does however have one major disadvantage: the sample size is not constant. Variable sample size designs are usually avoided by auditors because they dislike being in a situation where the number of observations is unpredictable at the planning stage; a sample size less than the target may lessen the belief in the audit, and a greater sample size may increase the cost. In this paper a new procedure, stabilised sieve sampling, which maintains the flexibility and simplicity of sieve sampling while overcoming the variable sample size problem, is proposed. Comparisons with unrestricted random, sieve and Lahiri sampling, carried out using the Stringer, cell and moment bounds by means of a simulation study based on two actual accounting populations with a range of error rates and amounts, yield favourable results. Stabilised sieve sampling has a higher coverage than the other methods. Its efficiency is similar to that of sieve sampling and it is consistently more efficient than unrestricted random and Lahiri sampling; the greatest gains in efficiency occur in large line item populations when the sample size is not small. [ABSTRACT FROM AUTHOR]