Restricting the complexity of species distribution models improves predictive ability

Allowing the response of species to ecological factors to be more flexible and complex improves the predictive capacity of ecological niche models, but if the sample is not large enough, problems of overfitting may arise. These problems can be reduced by using techniques that restrict the complexity of the model, such as penalized regression. ECOGESFOR researchers have successfully applied penalized logistic regression to distribution models of Spanish tree species. The penalized models outperform standard logistic regression models in predictive capacity and equal maximum entropy models (the current standard in species distribution models). Read article in Ecological modeling