Logistic Regression: A Self-learning Text, Third Edition (Statistics in the Health Sciences)

b. Sensitivity %¼100(98/114)¼86.0

Specificity %¼100(107/175)¼61.1

1 specificity %¼ 100 61.1¼100(68/175)¼39.9

False positive %¼100(68/166)¼41.0

False negative %¼100(16/123)¼13.0

c. The denominator for calculating 1specificity is

the number of true negatives (n 0 ¼175) whereas

the denominator for calculating the false positive

percentage is 166, the total number of patients who

were classified as postitive from the fitted model

using the cut-point of 0.300.

d. Correct (%)¼100(98þ107)/(114þ175)¼70.9,

which gives the percentage of all patients (i.e., cases

and noncases combined) that were correctly

classified as cases or noncases.

e. The sensitivity of 86.0% indicates that the model

does well in predicting cases among the true cases.

The specificity of 61.1 indicates that the model does

not do very well in predicting noncases among the

true noncases.

f. A drawback to assessing discrimination exclusively

using the cut-point of 0.300 is that the sensitivity

and specificity that results from a given cut-point

may vary with the cut-point chosen.

2. Plots for the following cut-points: 0.000, 0.200, 0.400,
   0.600, 0.800, and 1.000
   Sensitivity1.0

1–specificity

0.8

0.6

0.4

0.2

0.0

0.0 0.2 0.4 0.6 0.8 1.0

682 Test Answers