Are all patches worth exploring? Foraging desert birds do not rely on environmental indicators of seed abundance at small scales

Table 1 Spatio-temporal heterogeneity of seed removal

Model (~ predictors)	Residual d.f.	Residual deviance	Test	Δ deviance [d.f]	P
1. Full model (~ season * grid)
2. Main effects model (~ season + grid)	6	10.423	2 vs. 1	10.423 [6]	0.108
3. Temporal model (~ season)	8	124.506	3 vs. 2	114.083 [2]	< 0.001
4. Spatial model (~ grid)	9	83.665	4 vs. 2	73.242 [3]	< 0.001
5. Null model	11	191.038	5 vs. 2	180.614 [5]	< 0.001

Analysis of deviance of nested binomial Generalised Linear Models (GLMs; logit link) obtained through backwards elimination from the full model, with temporal (Season, four categories) and spatial (Grid, three categories) heterogeneity as predictors of the proportion of used devices in each experiment. The minimum adequate model includes both main effects (model 2, in italics)

ISSN: 1472-6785