Meta-analysis of a megafish: assessing patterns and predictors of Alligator Gar movement across multiple populations

Table 3 Comparison of models used to test water temperature-dependent relationships between Alligator Gar (Atractosteus spatula) displacement and changes in discharge in the Brazos and trinity rivers

Parameters	Model	K	logLik	AICc	ΔAICc	ω	Adj. R²
ΔQT*Study*	GAM	7	-2573.368	5161.2	0	0.995	0.18
ΔQT + Study*	GLM	6	-2580.652	5173.6	12.5	< 0.01	0.15
ΔQTStudy	GLM	9	-2577.538	5173.7	12.6	< 0.01	0.16
ΔQT + Study*	GAM	5	-2582.834	5175.9	14.7	< 0.01	0.14
ΔQ*T	GLM	5	-2584.214	5178.6	17.5	< 0.01	0.14
ΔQ + T	GAM	5	-2585.383	5182.4	21.2	< 0.01	0.13
ΔQ*T	GAM	4	-2586.976	5183.8	22.6	0	0.12
ΔQ + T	GLM	4	-2587.972	5184.1	22.9	0	0.12
ΔQ	GAM	3	-2589.96	5188.0	26.8	0	0.11
ΔQ	GLM	3	-2592.477	5191.0	29.9	0	0.10
T	GAM	3	-2600.295	5208.6	47.5	0	0.06
T	GLM	3	-2602.300	5210.7	49.5	0	0.05
Null	GLM	2	-2612.399	5228.8	67.7	0	0.00
Null	GAM	2	-2612.399	5228.8	67.7	0	0.00

Model parameters represent continuous measures of change in daily average discharge (ΔQ; z-score transformed) and daily average water temperature (T; °C) as well as a categorical factor representing the Brazos River or Trinity River mobile telemetry dataset (Study). Candidate model structures included generalized additive models (GAMs) for non-monotonic relationships or generalized linear models (GLMs) for monotonic relationships across seven possibilities: (1) null (intercept only), (2) fixed effect of T, (3) fixed effect of ΔQ, (4) two-way additive (+) modelling, (5) two-way interactive (*) modelling, (6) two-way interactive modelling and an additive effect, and (7) three-way interactive modelling. The table shows the number of parameters in the model (K), log-likelihood (LogLik), Akaike information criterion score adjusted for small sample size (AICc), the difference in AICc from the top model (ΔAICc), the Akaike weight (ω), and Adjusted R² values. Bolded text denotes the top model

ISSN: 2051-3933