Table 3.
Seasonal occupancy dynamics models following MacKenzie et al. (2006), applied to the guiña (Leopardus güigna), to define the base model structure for the subsequent model selection procedure to evaluate potential habitat configuration/quality and human–predator predictors. Fitted probability parameters are occupancy (ψ), colonisation (γ), extinction (ε) and detection (p). Models assess whether changes in occupancy do not occur (model 1.6), occur at random (models 1.5, 1.4) or follow a Markov Chain process (i.e. site occupancy status in a season is dependent on the previous season; models 1.0, 1.1, 1.2, 1.3). Initial occupancy (ψ1) refers to occupancy in the first of four seasons over which the güiña was surveyed. Model selection procedure is based on Akaike's Information Criterion (AIC). ∆AIC is the difference in AIC benchmarked against the best model, w i is the model weight, K the number of parameters and −2 × loglike is the value of the log likelihood at its maximum. The selected model is highlighted in bold
| Model | Seasonal dynamic models | ∆AIC | w i | K | −2 × loglike |
|---|---|---|---|---|---|
| 1.0 | ψ(.), γ(.), {ε= γ (1 − ψ)/ψ}, p(season) | 0.00 | 0.443 | 6 | 3,982.93 |
| 1.1 | ψ 1 (.), ε(season), γ(season), p (season) | 0.36 | 0.370 | 11 | 3,973.29 |
| 1.2 | ψ1(.), ε(.), γ(.), p(season) | 1.88 | 0.173 | 7 | 3,982.81 |
| 1.3 | ψ1(.), ε(.), γ(.), p(.) | 6.83 | 0.015 | 4 | 3,993.76 |
| 1.4 | ψ1(.), γ(.),{ε = 1 − γ}, p(season) | 41.78 | 0.000 | 6 | 4,024.71 |
| 1.5 | ψ1(.), γ(season),{ε = 1 − γ}, p(season) | 42.78 | 0.000 | 8 | 4,021.71 |
| 1.6 | ψ(.), {γ = ε = 0}, p(season) | 104.11 | 0.000 | 6 | 4,087.04 |