
102 NONPARAMETRIC GENERALIZED ADDITIVE REGRESSION
Predicted Value of PAD
Predicted Value of PAD
0.978
0.685
0.392
0.099
85.00
63.33
41.67
Age
BMI
20.00
15.00
19.33
23.67
28.00
85.00
63.33
41.67
Age
BMI
20.00
14
19
24
29
0.926
0.664
0.402
0.140
Predicted Value of PAD
0.851
0.639
0.427
0.215
LOESS (age) LOESS (BMI)
SPLINE (age, BMI)
SPLINE (age) SPLINE (BMI)
85.00
63.33
41.67
Age
BMI
20.00
14
19
24
29
As seen on the graphs, older patients with larger BMI have higher probability of be-
ing diagnosed with PAD. Based on the SAS output given below, the predicted proba-
bility P PAD that a 50-year-old patient with the BMI of 21.4 has PAD is 0.69749 for
the loess model, 0.61939 for the cubic spline