Do pseudo-absence selection strategies influence species distribution models and their predictions? An information-theoretic approach based on simulated data

If ecologists wish to build parsimonious GLM models that will allow them to make robust predictions, a reasonable approach is to use a large number of randomly selected pseudo-absences, and perform model selection based on an information theoretic approach. However, the resulting models can be expected to have limited fit.