Testing the model for one predator and two mutualistic prey species

TH Fay
JC Greeff
BE Eisenberg
HT Groeneveld

Abstract

With increasing frequency, models are being used to synthesize information and predict species response to ecosystem changes. Since it is reasonable to assume that models will continue to play a role in ecological and environmental studies, there is a need for a means of assessing model performance. The traditional approach of validation in terms of goodness-of-fit is such a process, but does not guarantee the "correctness" of model predictions. To increase confidence in the predictive qualities of the model, additional validation techniques are investigated, to ensure the effective and appropriate use of the model in the decision making processes in managing wildlife undertakings. An initial graphical investigation is executed, to obtain a qualitative indication of model robustness, and is supplemented with quantitative techniques, such as model efficiency and Monte Carlo hypothesis testing, before conclusions are drawn with respect to model suitability.