Statistical Evaluation of Continuous-Scale Diagnostic Tests with Missing Data

摘要: The receiver operating characteristic (ROC) curve methodology is the statistical for assessment of accuracy diagnostics tests or bio-markers. Currently most widely used methods inferences ROC curves are complete-data based parametric, semi-parametric nonparametric methods. However, these cannot be in diagnostic applications with missing data. In practical situations, data occur more commonly due to various reasons such as medical being too expensive, time consuming invasive. This dissertation aims develop new evaluating biomarkers presence Specifically, novel will developed different types (i) inference area under (AUC, which a summary index test) and (ii) joint sensitivity specificity continuous-scale test. this dissertation, we provide general framework that combines empirical likelihood estimation equations nuisance parameters proposed have sound theoretical properties. development challenging because profile log-empirical ratio statistics not standard sum independent random variables. power approaches jackknife method studies. Therefore, they expected robust, accurate less computationally intensive than existing evaluation competing tests. INDEXWORDS: AUC, Bootstrap, Diagnostic tests, Empirical likelihood, Estimating equations, Imputation, Jackknife, Missing data, curve, Sensitivity, Specificity, Verification bias STATISTICAL EVALUATION OF CONTINUOUS-SCALE DIAGNOSTIC TESTS WITH MISSING DATA