Male dominance linked to size and age, but not to 'good genes' in brown trout (Salmo trutta)

Table 1 The Influence of paternal, maternal and paternal × maternal interaction effects on embryo mortality in the 1^st experiment.

		Model parameters						Likelihood ratio tests (LRT) with reference model
Model	Effect tested	Random	Fixed	Number (k)	ln L	AIC	ΔAIC	χ²	d.f.	P
reference model		F, M	T	3	-280.19	568.38
full model	Male × Female	F, M, F × M	T	4	-280.19	570.38	2.00	0.00	1	1
female model	Male	F	T	2	-283.04	572.07	3.69	5.69	1	0.017
male model	Female	M	T	2	-403.65	813.30	244.92	246.92	1	<0.0001

Four logistic mixed effect models are compared to test if male (M), female (F), and male × female interaction (M × F) effects explain a significant part of the variance in embryo mortality (a binary response variable; egg number n = 2028). The random, fixed, and total number (k) of parameters are given for every model. The goodness of fit is given by the logarithm of the approximated likelihood (ln L) and the Akaikes information criterion (AIC). A measure to compare the quality of fit between two models is the difference of AICs (ΔAIC) between two models. The reference model explains our data best as the more complex full model does not significantly improve the qualtiy of fit (see ΔAIC; LRT), i.e. the male × female interaction effect did not explain a significant part of the variance in embryo mortality. The table therefore gives the differences in AICs between the reference model and the other models. Furthermore, likelihood ratio tests (LRT) between the reference model and the other models are given to test which parameter significantly improves the goodness of fit.

ISSN: 2730-7182