A Simulation Study of the Performance of Parametric Dollar Unit Sampling Statistical Procedures
通过模拟研究评估多种参数化美元单位抽样方法，这些方法能结合先验信息与样本信息计算总误差上限，旨在解决非参数方法处理低估和中等高错误率下风险过大的问题。
Two major reservations concerning dollar unit sampling were summarized by Kaplan [1975b, p. 141] thus: (1) treatment of understatements with nonparametric methods is considered by some to be ad hoc (this concern relates to the computation of bounds on net error and has been identified in a dollar unit sampling conference as a primary research issue of dollar unit sampling-see Felix, Leslie, and Neter [1982, p. 100]); and (2) there is concern that the a risk (rejecting materially correct populations) may be excessive in moderate to high error rates for the sample sizes normally used in practice (this has been characterized as the efficiency of an audit procedure in AICPA SAS No. 39 [1981, sec. 13]). The purpose of this paper is to evaluate the use of various parametric dollar unit sampling methods designed to address these concerns. Those considered here have the additional advantage that they allow incorporation of prior information as well as sample information in developing a bound on total error, since the methods are Bayesian. The combination of these potential advantages suggests that Bayesian parametric dollar unit sampling models may be an important alternative set of models that