Slide 7: Model evaluation
The DBF (or database) file representing the species distribution will then be imported into Arcview, and displayed as an event theme, as you did with the full dataset. You can evaluate the accuracy of the model by comparing the known and the modelled distributions of the species within ArcView. You may well have to play with the legend of the two layers in order to ascertain which of your models is more accurate. Another means of evaluating the accuracy of the models (and the one that should inform your final decision) is through receiver operating character analysis (which was previously discussed in the modelling chapter). The ROC can be calculated within the statistical package, and is displayed as a value called AUC. This stands for the Area Under the Curve, and is a value between 0 and 1. Values over 0.8 are considered to be good models, but any value over 0.75 can be used if that is the maximum accuracy you are able to obtain.