Abstract:
|
Youden's Index is one statistic used to quantify and compare the performance of diagnostic tests in two class testing scenarios. Common methods used to create confidence intervals and tests of hypotheses regarding Youden's Index includes the delta and generalized methods as well as resampling (bootstrap) methods. When prevalence or cost of misclassifications vary, the Generalized Youden's Index is used which is a weighted function (weighted by the products of prevalence and costs) of the correct classification rates of the diagnostic test, however, as a simplifying assumption, prevalence and cost is typically fixed. This work examines inference for Youden's Index treating the prevalence as a random variable. Confidence intervals will be developed using different distributional assumptions for the prevalence, such as Uniform or Beta, and for the classes, such as Normal or empirically estimated. Resampling methods and simulation will be the primary tool for development of the confidence interval. Application of the methods to a case study of epidemiological data will also be demonstrated.
|