DISCUSSION OF Aggregation of Evidence in Auditing: A Likelihood Perspective.
从似然视角探讨审计中多变量证据网络的聚合框架，通过敏感性分析验证模型结构变化对输出的影响，帮助审计人员判断模型逻辑一致性。
Abstract This article focuses on the general framework for aggregating evidence from a network of variables using the likelihood perspective. There are several ways in which models such as the one proposed by researchers can be validated. One validation technique is the use of sensitivity analysis. For evidential networks, sensitivity analysis can be performed by computing the output and model structure. Sensitivity analysis is important and useful for at least two reasons. First, it is mathematical technique for validating analytical models, particularly when alternative methods are less feasible. Second, evidential networks in auditing may be improperly specified. According to the author, for the sake of clarity and computational tractability sensitive analysis is performed primarily on the model presented by researchers. Although the implications from the analysis are relevant to the aggregation scheme proposed by researchers in general. The first part of the analysis examined the impact on the output of the model of varying the model structure. The second part is a more basic form of sensitivity analysis. It consists of constructing simple numerical inputs into the model and verifying that the output of the model is logically consistent.