Ecological Archives A022-122-A3
Amanda E. Martin, Lenore Fahrig. 2012. Measuring and selecting scales of effect for landscape predictors in species–habitat models. Ecological Applications 22:2277–2292. http://dx.doi.org/10.1890/11-2224.1
Appendix C. Comparisons of predictive performance during model extrapolation for the best single-scale and multi-scale models for each species and each statistical framework.
Table C1. Comparisons of predictive performance during model extrapolation for the best single-scale (SS) and multi-scale (MS) occupancy (OCC), generalized linear (GLZ), generalized additive (GAM) and classification and regression tree (CART) models for each species, where all landscape predictors were measured at the same scale in SS models and predictors were each measured at their individual scale of effect in MS models. Predictive performance during extrapolation was estimated by using the models parameterized with the training sites to make predictions for the extreme test sites and comparing the area under the receiver operating characteristic curve between SS and MS models for a given species and framework using a bootstrap method, with 5000 samples. For OCC, GLZ, and GAM we used model averaged predictions, including all models for which the change in the Akaike Information Criterion (corrected for small samples) from the most supported model was less than or equal to two.
[Back to A022-122]