A Dynamic Modeling Framework for Predicting Species Occurrences: Models for Managers Author: Jason W. Karl P.O. Box 298 Phone: 509-996-2490 A challenge in species occurrence modeling is to communicate the dynamic nature of modeling (i.e., there is no "right" model). Selection of a threshold value for determining suitable/unsuitable habitat in statistical models will affect rates of omission (species detected where not predicted) and commission (species not detected where predicted) error. Different applications will determine relative costs of error types. We present a GUI for ArcView GIS that allows users to manipulate threshold values of logistic regression models, evaluate error changes, and display the model that best suits their needs. We demonstrate this tool with models for several forest songbirds in north Idaho. |