Appendix 2. Summary of Fitted Generalized Additive Models.( Predictors: Log Area, Distribution

Appendix 2. Summary of fitted Generalized Additive Models.( Predictors: log Area, Distribution: normal)

The Response fields specify, respectively,the name of response (dependent) variable.

Type specifies (for one predictor or two predictor, separated by "+" character) its complexity within the model: lin means that a (generalized) linear term was chosen during stepwise selection, while describes a smooth term with complexity value 2 measured in degrees of freedom best model was selected by Akaike Information Criterion (AIC) values.

R2[%] provides a measure of explained variation, paralleling the coefficient of determination in classical regression, calculated here as the ratio of the deviance explained by the fitted model and the deviance of a null model (with no predictors), multiplied by 100

F test statistic and following p estimate of type I error rate correspond to an overall parametric test of the selected model against the null model.

Response / Type / R2[%] / F / p
K / s2 / 3.3 / 3.7 / 0.02649
X1 / s2 / 8.7 / 10.3 / 0.00006
X3 / s2 / 2.7 / 3 / 0.05159
ABCD / s2 / 15.7 / 20 / <0.00001
F / lin / 3.8 / 8.5 / 0.00379
H1 / lin / 2.6 / 5.7 / 0.01725
J / lin / 2.2 / 4.8 / 0.02997
LO / lin / 1.2 / 2.6 / 0.11127
M / lin / 1 / 2.2 / 0.14047
N / s2 / 5.3 / 6.1 / 0.00274
SN / s2 / 8.2 / 9.6 / 0.0001
TC / s2 / 7.3 / 8.5 / 0.00027
MP / s2 / 10.4 / 12.4 / <0.00001
W1 / s2 / 20 / 26.8 / <0.00001
W2 / s2 / 11.2 / 13.6 / <0.00001
WS / s2 / 2.8 / 3.1 / 0.04705
TD / lin / 1.8 / 4 / 0.04739
X2 / lin / 5.9 / 13.4 / 0.0003