(a) (b) (c)

GAM models with different degrees of freedom. (a) df=3. (b) df=5. (c) df=7.

e 4.16 shows this process through varying the degrees of freedom

to 20. It can be seen that the model performance was the best

degrees of freedom was eight and the AIC value was minimised

.

4.16. GAM model optimisation using AIC for the data in Figure 4.15.

tatistics of a GAM model can also be calculated by calling the

y function, for which the input is a model constructed by the gam

By calling the coef function with a constructed GAM model as

a quantitative GAM for the data model can be derived. One such

model constructed for the data shown in Figure 4.15 was

ed as below,

y = 0.8379+ 0.2713s(x) +