Bayesian Bounds for Monetary Unit Sampling in Accounting and Auditing
扩展了货币单位抽样中总体错误金额的贝叶斯界限研究，通过分析Cox和Snell界限的特性、开发替代贝叶斯模型并模拟评估其稳健性，仅关注高估错误。
Monetary unit sampling (MUS) has become widely used by auditors. With MUS, individual monetary (e.g., dollar) units are selected at random and the line item to which a monetary unit belongs is then audited. In this way line items with large dollar values have greater probability of being sampled than do line items with small dollar values. It is the purpose of this paper to extend the limited research performed to date on constructing Bayesian bounds for the population total error amount when monetary unit sampling is utilized by (1) studying the characteristics of the Bayesian bound proposed by Cox and Snell [1979], (2) developing alternative Bayesian models in which the Cox and Snell assumptions are modified to see how sensitive the bounds are to these assumptions, and (3) conducting a simulation study to assess the robustness of the Cox and Snell bound to a wide variety of population conditions. We do not consider in this paper errors of understatement; we confine the discussion entirely to errors of overstatement. The Bayesian approach can be directly generalized to consider both overstatement and understatement errors, but additional computational complexities may be encountered. Also, monetary unit sampling may not be the most effective type of sampling when understatement errors play a dominant role. We