Table 1. Model selection results from the avian attack and non-predator disturbance analysis.
| Model | K | LL | AICc | Evidence ratio | AICc wi |
| Avian Attacks | |||||
| Color+Blotching+Region+Region*Color+Region*Blotching | 7 | 13.735 | 28.006 | 14.731 | 0.041 |
| Color+Blotching+Color*Blotching+Region+Region*Color | 7 | 12.859 | 27.131 | 9.509 | 0.063 |
| Color+Region+Section[Region]+Region*Color | 6 | 14.192 | 26.395 | 6.582 | 0.091 |
| Color+Blotching+Region+Region*Color | 6 | 12.680 | 24.883 | 3.091 | 0.193 |
| Color+Region+Region*Color | 5 | 12.482 | 22.626 | 1.000 | 0.597 |
| Non-predator Disturbances | |||||
| Blotching | 3 | 2.023 | 8.081 | 1.000 | 0.999 |
Only the significant models (P<0.05) which demonstrated some level of potential importance according to the number of parameters (K), log likelihood (LL), Akaiki Information Criteria corrected for small sample size (AICc), evidence ratio, and AICc weight (wi) are presented. In the column presenting the models an asterisk (*) indicates that parameters are crossed and brackets ([…]) indicate that the preceding parameter is nested within the bracketed parameter. The model containing Color+Region+Region*Color appears to be favored for avian attacks, while the model containing Blotching is favored for non-predator disturbances.