Statistical Methods - Text Mining Techniques for Healthcare Provider Quality Determination - page 100

Table 15. Classification table for logistic regression with pneumonia and septicemia

Classification Table

Prob
Level

Correct Incorrect Pe rcentages

Event Non-Event Event Non-Event Correct Sensi-

tivity

Speci-
ficity

False
POS

False
NEG

0.580

782E4 0 167E3 0 97.9 100.0 0.0 2.1 .

0.600 78E5 9948 157E3 25175 97.7 99.7 6.0 2.0 71.7

0.620 78E5 9948 157E3 25175 97.7 99.7 6.0 2.0 71.7

0.640 78E5 9948 157E3 25175 97.7 99.7 6.0 2.0 71.7

0.660 78E5 9948 157E3 25175 97.7 99.7 6.0 2.0 71.7

0.680 78E5 9948 157E3 25175 97.7 99.7 6.0 2.0 71.7

0.700 78E5 9948 157E3 25175 97.7 99.7 6.0 2.0 71.7

0.720 78E5 9948 157E3 25175 97.7 99.7 6.0 2.0 71.7

0.740 78E5 9948 157E3 25175 97.7 99.7 6.0 2.0 71.7

0.760

78E5 9948 157E3 25175 97.7 99.7 6.0 2.0 71.7

0.780 78E5 9948 157E3 25175 97.7 99.7 6.0 2.0 71.7

0.800

78E5 9948 157E3 25175 97.7 99.7 6.0 2.0 71.7

0.820 78E5 9948 157E3 25175 97.7 99.7 6.0 2.0 71.7

0.840 768E4 41608 126E3 149E3 96.6 98.1 24.9 1.6 78.1

0.860 768E4 41608 126E3 149E3 96.6 98.1 24.9 1.6 78.1

0.880 768E4 41608 126E3 149E3 96.6 98.1 24.9 1.6 78.1

0.900 768E4 41608 126E3 149E3 96.6 98.1 24.9 1.6 78.1

0.920 768E4 41608 126E3 149E3 96.6 98.1 24.9 1.6 78.1

0.940 768E4 41608 126E3 149E3 96.6 98.1 24.9 1.6 78.1

0.960 731E4 63391 104E3 517E3 92.2 93.4 37.9 1.4 89.1

0.980 731E4 63391 104E3 517E3 92.2 93.4 37.9 1.4 89.1

1.000 0 167E3 0 782E4 2.1 0.0 100.0 . 97.9

( ˆ )= b

gY X

where g() is a non-linear link function. When the response variable follows a normal distribution, the
link function is the identity. This identity link explains how this model generalizes the linear model.
Examples of link functions for a variety of probability distributions are given in Table 17.
The generalized linear mixed model extends the generalized model when there is a random effect
in the data. It also generalizes the linear mixed model when the dependent variable is not normally
distributed. However, the random effects must be normal. The generalized linear mixed model has the
following form

gEYu (( /))=+ X u b