Plenary Talk

Title: Sequential Methods for Evaluation of Medical Tests
Presenter: Dr. Aiyi Liu, National Institute of Child Health and Human Development, National Institutes of Health, USA
Abstract:

Abstract: Medical tests are usually applied to identify human subjects with certain disease conditions. The performance of a medical test relies on its sensitivity, i.e. the probability of correctly identifying a diseased subject, and specificity, i.e. the probability of correctly identifying a non-diseased subject. When the test yields continuous outcomes, the receiver operating characteristic curve and the area under the curve are common tools to evaluate the performance of the test.

Aiming at reducing study costs and putting efficient diagnostic tests into practice in a timely manner, this talk will present a number of sequential approaches recently developed by the author and his collaborators to assess the performance of diagnostic tests. For a single diagnostic test with binary outcomes, exact sequential methods, two-stage procedures in particular, are proposed to evaluate the test's sensitivity and specificity. For comparison of two diagnostic tests with continuous outcomes, the parametric method based on normally transformed data is presented, along with a non-parametric method based on the Mann-Whitney statistics.