Advances and Applications in Statistics
Volume 2, Issue 3, Pages 301 - 309
(December 2002)
|
|
OPTIMAL PREDICTION FOR DIAGNOSTIC TESTS USING LOGISTIC REGRESSION ANALYSIS: AN ALGORITHM AND LARGE SAMPLE PROPERTIES
George Tzavelas (Greece) and Demosthenes B. Panagiotakos (Greece)
|
Abstract: Until now day’s receiver
operating characteristic curves (ROC) were,
mainly, applied as the cut off criterion for
indicating a positive diagnostic test. This
arbitrary criterion is often used in making a
choice between competing tests, although the
procedure takes no account of the frequency
distribution of the disease being tested for.
Moreover, other cutoff-value methods for
classifying sample and future observations are
linked to discriminant analysis and specially to
Fisher’s linear discriminant function, which
could not be generalized because it is produced
by assumptions which do not hold in the logistic
regression, which is considered essential in
diagnostic evaluations. The aim of this work is
to provide an algorithm for the detection of the
optimal cut off point of a diagnostic criterion
using the methodological background of logistic
regression analysis. |
Keywords and phrases: cut off point, logistic regression. |
|
Number of Downloads: 6 | Number of Views: 1209 |
|