Table 2.

Summary of statistics and results from the model fitting and selection procedure for biomass for each size class and mortality scenario

Size class	Model	P value	Estimated df*	GCV†	AIC†	Adjusted R2
(A) Adult mortality
Juveniles	Smoother	.004	2.69	3.16	250	0.36
	Linear	.017		4.19	259	0.10
Nonrecruits	Smoother‡	.082	2.24	4.37	283	0.20
	Linear	.971		5.43	290	−0.03
Adults	Smoother	<.001	1	11.4	349	0.59
	Linear	<.001		11.4	350	0.59
Total	Smoother	<.001	3.73	0.03	367	0.53
	Linear	<.001		0.03	375	0.40
(B) Juvenile mortality
Juveniles	Smoother	.001	4.09	1.72	226	0.48
	Linear	<.001		2.1	232	0.31
Nonrecruits	Smoother	.098	1	0.02	249	0.26
	Linear	.001		0.02	249	0.26
Adults	Smoother	.209	3.31	17.5	368	0.12
	Linear	.571		18.62	371	−0.02
Total	Smoother	.380	1.16	17.18	378	−0.03
	Linear	.793		17.19	378	−0.03

Statistics for selected models are given in bold. P values are for smoothing terms and slopes for GAM and linear regression models, respectively.

*Estimated df for the smoothing term as provided by the fitting procedure. These should not be to close to 1 or the maximum df possible (here 6).

^†GCV scores are for GAMs with and without (= linear) smoothing terms for mortality rate, whereas AICs are for GAMs with smoothers and standard linear models (= linear). The lower the GCV score, the better the model performance. GAMs without smoothing terms are here essentially standard linear models, so the both comparisons select the same models.

^‡Marginally significant.