Table 2.
Selection of the best model.
| Model I | Model II | Model III | Model IV | |||||
|---|---|---|---|---|---|---|---|---|
| χ2 | P | χ2 | P | χ2 | P | χ2 | P | |
| Both years | 5.85 | 0.12 | 21.47 | < 0.0001 | 29.54 | < 0.0001 | 65.17 | < 0.0001 |
| 2017 | 3.39 | 0.33 | 13.65 | 0.003 | 20.59 | 0.0001 | 39.42 | < 0.0001 |
| 2018 | 3.13 | 0.37 | 8.58 | 0.035 | 9.69 | 0.02 | 26.69 | < 0.0001 |
| 2017 M600 | 1.79 | 0.62 | 6.16 | 0.104 | 11.17 | 0.01 | 19.38 | 0.0002 |
| 2017 B400 | 2.24 | 0.52 | 7.98 | 0.046 | 9.89 | 0.02 | 20.46 | 0.0001 |
| 2018 M600 | 4.98 | 0.083 | 6.89 | 0.032 | 4.23 | 0.120 | 14.48 | 0.0007 |
| 2018 M1000 | 4.24 | 0.236 | 7.75 | 0.051 | 21.88 | < 0.0001 | 19.03 | 0.0002 |
Observed transition probabilities were compared against predicted values for each of the four models of patch attractiveness (defined in the text). The Chi-square value and associated probability, P value, are presented here. A model with a probability > 0.05 (bolded) indicates a good fit to the empirical data.