论文部分内容阅读
[目的]探讨无金标准条件下诊断试验准确性评价的潜分类方法。[方法]介绍潜分类模型在无金标准诊断试验评价中的原理、试验设计和评价方法,用两人群两试验实例说明潜分类方法的应用。[结果]对于二分类反应变量,假设条件独立和试验准确性稳定,至少需要两个人群两种试验方法或一个人群三种试验方法才能满足模型可识别性并用于频率学派统计评价;贝叶斯统计不需满足模型的可识别性,但需引入先验分布,且存在先验依赖性。[结论]潜分类方法可用于无金标准时的诊断试验评价,但要选择适合的试验设计和评价方法。
[Objective] To explore the potential classification method for evaluating the accuracy of diagnostic tests without gold standard conditions. [Methods] The principle, experimental design and evaluation methods of the sub-taxonomic model in the evaluation of non-gold standard diagnostic tests were introduced. The application of the sub-classification method was illustrated with two experimental examples of two populations. [Results] For the dichotomous response variables, assuming independent conditions and stable experimental accuracy, at least two test methods of two populations or three test methods of one population are needed to satisfy the model identifiability and be used in the statistical evaluation of frequency school. Bayes Statistics do not need to satisfy the identifiability of the model, but need to introduce a priori distribution, and there is an a priori dependency. [Conclusion] The sub-classification method can be used to evaluate the diagnostic test without gold standard, but choose the appropriate experimental design and evaluation method.