Note, however, if two models that are not
nested are compared, there is no guarantee
which model will have a larger AUC. For exam-
ple, the model that contains only the three vari-
ables that were dropped, namely HEAD,
AGECAT, and FLEX has an AUC of 0.660,
which is smaller than the AUC of 0.731
obtained for the two variable (reduced) model
involving WEIGHT and PATELLAR.EXAMPLE (continued)
Model 3: HEAD, AGECAT, and
FLEX
AUCModel 3 ¼ 0 : 660
Reduced Model: WEIGHT and
PATELLAR
AUCModel 2 ¼ 0 : 731
Thus, Model 2 (fewer variables)
discriminates better than
Model 3 (more variables)370 10. Assessing Discriminatory Performance of a Binary Logistic Model