ABSTRACT
Screening and diagnosis are encountered frequently in medicine. Statistical methods are needed to rigorously evaluate the diagnostic accuracy, usually expressed as numerical probabilities. Sensible accuracy measures should reflect how often the test can correctly identify the disease status of a subject. Accuracy analysis may serve as an alternative approach to study the association between survival outcomes and predictors, taking a rather different angle
from hazard regression. Statistical tools such as the ROC (receiver operating characteristics) analysis are introduced in this chapter.
