TABLE 3.
Results of the broken‐stick regression for the boosted regression trees models considering the original distribution model, a rat‐free model (where we exchanged the variable distance to coast with distant to the nearest ret‐free island) and the rat‐invaded model (where we replaced distance to coast by distance to nearest rat‐invaded island)
| Species | Models | Estimated break point (km) | Confidence Intervals (km) | |
|---|---|---|---|---|
| Lower | Upper | |||
| Red‐footed booby | Distribution Model | 37.4 | 34.0 | 40.8 |
| Rat‐invaded Model | 60.0 | 55.0 | 65.0 | |
| Rat‐free Model | 37.8 | 34.9 | 40.7 | |
| Brown noddy | Distribution Model | 49.0 | 45.6 | 52.3 |
| Rat‐invaded Model | 56.6 | 51.9 | 61.3 | |
| Rat‐free Model | 48.6 | 46.6 | 50.6 | |
| White tern | Distribution Model | 27.1 | 26.2 | 27.9 |
| Rat‐invaded Model | 24.4 | 23.4 | 25.4 | |
| Rat‐free Model | 54.5 | 50.9 | 58.0 | |
| Wedge‐tailed shearwater | Distribution Model | 91.2 | 57.1 | 125.4 |
| Rat‐invaded Model | 32.1 | 30.4 | 33.8 | |
| 64.0 | 61.4 | 66.7 | ||
| Rat‐free Model | 21.2 | 18.9 | 23.5 | |
| 83.3 | 76.3 | 90.3 | ||
The broken‐stick regression is represented by the breaking point which is the inflection point of the partial dependence plot from each model. Lower and upper confidence interval at 95% is reported in each break point.