Introduction In this chapter, we describe and illustrate methods for
assessing the extent that a fitted binary logistic model can
be used to distinguish the observed cases (Y¼1) from the
observed noncases (Y¼0).
One approach for assessing such discriminatory perfor-
mance involves using the fitted model to predict which
study subjects will be cases and which will not be cases
and then determine the proportions of observed cases and
noncases that are correctly predicted. These proportions
are generally referred to as sensitivity and specificity
parameters.Another approach involves plotting a receiver operating
curve(ROC) for the fitted model and computing the area
under the curve as a measure of discriminatory perfor-
mance. The use of ROCs has become popular in recent
years because of the availability of computer software to
conveniently produce such a curve as well as compute the
area under the curve.Abbreviated Outline
Outline
The outline below gives the user a preview of the material
to be covered by the presentation. A detailed outline for
review purposes follows the presentation.I. Overview (pages 348–350)
II. Assessing discriminatory performance using
sensitivity and specificity parameters (pages
350–354)
III. Receiver operating characteristic (ROC) curves
(pages 354–358)
IV. Computing the area under the ROC: AUC (pages
358–365)
V. Example from study on screening for knee
fracture (pages 365–370)
VI. Summary (page 371)346 10. Assessing Discriminatory Performance of a Binary Logistic Model