Table 1 Model selection of the two models based on likelihood ratio tests.

Model		Fixed terms	Random terms	Test	AIC	AICc	BIC	logLik	deviance	Chisq	df	p
1	1a	Habitat + Habitat * Phase			8,465	8,473	8,726	−4,176	8,353
	1b	Habitat + Habitat * Phase	Habitat/ID	1a vs 1b	5,738	5,759	6,130	−2,785	5,570	2,782	28	<0.001
	1c		Habitat/ID		7,851	7,854	7,987	−3,897	7,793
	1d	Habitat	Habitat/ID	1c vs 1d	7,769	7,799	7,959	−3,863	7,726	67	6	<0.001
	1e	*Habitat + Habitat Phase**	Habitat/ID	1d vs 1e	5,738	5,759	6,130	−2,785	5,570	2,156	49	<0.001
2	2a	Behaviour * Habitat + Behaviour Phase + Behaviour Phase * Habitat	Behaviour/ID		12,745	12,772	13,706	−6,205	12,409
	2b	Behaviour * Habitat + Behaviour Phase + Behaviour Phase * Habitat	Behaviour/ID	2a vs 2b	12,498	12,527	13,493	−6,075	12,150	259	6	<0.001
	2c		Behaviour/ID		81,124	81,083	81,164	−40,555	81,110
	2d	Behaviour	Behaviour/ID		81,083	81,124	81,134	−40,532	81,065	46	2	<0.001
	2e	Behaviour * Habitat	Behaviour/ID	2c vs 2d	80,272	80,273	80,444	−40,106	80,212	852	21	<0.001
	2f	Behaviour * Habitat + Behaviour * Phase	Behaviour/ID	2d vs 2e	14,439	14,441	14,713	−7,172	14,343	65,869	18	<0.001
	2g	*Behaviour Habitat + Behaviour Phase + Behaviour Phase * Habitat**	Behaviour/ID	2e vs 2f	12,498	12,527	13,493	−6,075	12,150	2,193	126	<0.001

First the random terms were selected keeping fixed terms the same, and secondly the fixed terms were defined with the best selection of random terms. The model highlighted in bold is the final ‘best’ fit model for that analysis (1e and 2g respectively). Both models resulted in a GLMM model with a random slope. Stage = breeding stage, AIC = Akaike information criterion, AICc = corrected Akaike’s information criterium, BIC = Bayesian information criterion, logLik = log-likelihood ratio statistic, Chisq = Chi-square statistic, df = degrees of freedom. The significance level was set at α = 0.05.

Quick links

Search