Abstract：
For low prevalence disease, we consider estimation of the odds ratio for two specified groups of individuals using group testing data. Broadly the two groups may be classified as \"the exposed\" and \"the unexposed.\" Often in observational studies, the exposure status is not correctly recorded. In addition, diagnostic tests are rarely completely accurate. The proposed model accounts for imperfect sensitivity and specificity of diagnostic tests along with the misclassification in the exposure status. For model identifiability, we make use of internal validation data, where a subsample of reasonably small size is selected from the original sample by simple random sampling without replacement. Pseudo-maximum likelihood method is employed for the estimation of the model parameters. The performance of group testing methodology is compared with individual testing for different parametric configurations. A limited data study related to COVID-19 prevalence is performed to illustrate the methodology.
摘要：
对于低患病率疾病，我们考虑使用分组测试数据估计两组特定个体的比值比.从广义上讲，这两组可以分为“暴露者”和“未暴露者”。“通常在观察性研究中，未正确记录曝光状态。此外,诊断测试很少是完全准确的。所提出的模型考虑了诊断测试的敏感性和特异性不完善以及暴露状态的错误分类。对于模型的可识别性，我们利用内部验证数据，其中，通过简单随机抽样从原始样本中选择一个规模相当小的子样本，无需替换。采用伪最大似然法估计模型参数。将组测试方法的性能与针对不同参数配置的单个测试进行比较。进行了与COVID-19患病率相关的有限数据研究来说明该方法。
