Feeding selectivity and a functional trade-off in a benthic fish with a continuous morphological variation: an experimental test

Table 2 Results of model selection in the fits of GLMs for feeding efficiency (as EA)

[Model]:		Coef.	t	P	AIC	ΔAIC
[1–1]: Efficiency (EA) ~ SL + MI1	SL	0.001	3.76	0.001	-71.08	-2.79
	MI1	0.04	2.05	0.054	-71.08
[1–2]: Efficiency (EA) ~ SL + MI1²	SL	0.002	3.72	0.001	-68.29
	MI1²	-0.05	-1.2	0.24	-68.29
[2−1]: Efficiency (EA) ~ SL + MI2	SL	0.001	3.56	0.002	-68.69	-0.81
	MI2	0.02	1.35	0.19	-68.69
[2–2]: Efficiency (EA) ~ SL + MI2²	SL	0.002	3.63	0.002	-67.88
	MI2²	-0.02	-1.03	0.31	-67.88
[3−1]: Efficiency (EA) ~ SL + MI3	SL	0.001	3.39	0.003	-66.68	6.48
	MI3	0.001	0.05	0.96	-66.68
[3−2]: Efficiency (EA) ~ SL + MI3²	SL	0.001	2.31	0.03	-73.16
	MI3²	0.12	2.55	0.02	-73.16

The best model for each MI was determined based on Akaike information criterion (AIC). ΔAIC was calculated as the difference between the AIC of the linear model minus the AIC of the quadratic model. Coefficients (Coef.) indicate regression estimate values, and the t values are the statistics in the GLM Gaussian models. A ΔAIC larger than 4 indicates more support for a quadratic model, while a ΔAIC less than − 4 indicates more support for a linear model. A ΔAIC between − 4 and 4 suggests equivalent support between them [33]

ISSN: 2056-3132