Table 3. Effect of covariatesa on occupancy ().
Modelb | ΔAIC | AIC weight | Number of parameters | Deviance (-2 Log-likelihood) |
ψ (B × WildP) | 0 | 0.654 | 12 | 837.02 |
ψ (B × Dist) | 3.05 | 0.142 | 12 | 840.07 |
ψ (WildP) | 4.03 | 0.087 | 10 | 845.05 |
ψ (B × Hab) | 4.18 | 0.081 | 12 | 841.2 |
ψ (B) | 6.32 | 0.027 | 10 | 847.34 |
ψ (B × PrincipP) | 9.81 | 0.004 | 12 | 846.82 |
ψ (Hab) | 12 | 0.001 | 10 | 853.02 |
ψ (Dist) | 14.69 | 0.0004 | 10 | 855.71 |
ψ (PrincipP) | 18.97 | 0 | 10 | 859.99 |
Note: Model rankings are based on Akaike’s Information Criterion (AIC).
Covariates used to model detection probability were Block (B), Wild prey index (WildP), Principal prey index (PrincipP), Disturbance (Dist) and proportional habitat per cell (Hab).
In all models the probability of detection () was modelled as ‘B + Substrate’ based on model selection results presented in Table 1. Segment-level occupancy parameters ( and ) were modelled on ‘B’ (Block). ‘×’ denotes covariates were modelled as an interaction.