Table 4.

Comparison of the predictive abilities of the functional models Functional Generalized Linear Model (FGLM), Functional Generalized Spectral Additive Model (FGSAM), and Functional Generalized Kernel Additive Model (FGKAM) ^h.

Goodness-of-Fit Measures of the Functional Models	FGKAM	FGLM	FGSAM
Adjusted R-sq (%)	65.70	57.90	67.30
Dev. Explained (%)	75.10	61.20	72.40
MSPE	7.52	7.50	11.38
Pred. coverage (%)	90.00	95.00	92.50

^h Here, we display some goodness-of-fit measures R-sq(adj), Mean Square Prediction Error (MSPE), and predictive coverage as a tool to compare the functional models FGLM, GGSAM, and FGKAM. In terms of the predictive abilities, all models performed well, none did better than the others. FGSAM fit the best with adjusted R-sq (67.3%), but FGLM had the best predictive coverage (95%) and FGSAM obtained the best explained deviance (75.1%).