Title:
Discriminant Analyses in Biomedical Studies
Abstract:
The discriminant analyses in biomedical research are in general associated with discriminating diseased individuals from normal healthy individuals
through estimating the diagnostic accuracy. The diagnostic accuracy of a given diagnostic test has been traditionally measured by sensitivity and
specificity, as well as the area associated with the Receiver Operating Characteristic (ROC) curve. In this talk, we will first discuss several
extensions to the statistical estimation/comparison of diagnostic accuracy based on ROC curves with different types of diagnostic tests. We will
then
discuss the problem of measuring/estimating/comparing diagnostic accuracy when more than two diagnostic alternatives are available. Finally, we will
present an application of the ROC curve/surface-based statistical methodologies to search for a subset of genes whose microarray expression profiles
provide (in certain sense) the optimum discriminatory power between two sources of RNA.