Slide 11: Response curves estimation of different models
Here are some examples of response curves assumed or derived by various modelling procedures. The simplest form of modelling, Generalised Linear Modelling uses linear, Gaussian (binomial) response curves.
Generalised Additive Modelling (GAM), uses iterative (repeated) functions to derive more complex, polynomial fitted lines. Tree Models operate through a series of binary Yes-No decisions to decide whether conditions are appropriate, so strictly speaking the model does not have a response curve.
Envelope, or bioclimatic modelling in its simplest form, has a similar binary structure, with presence assumed at the 100% level within the response bounds, and absence assumed without.
CCA (or ordination) uses simple Gaussian distributions, but works by "clumping" data into standard deviations from the centre point, giving graduated response classes instead of outright likelihoods.
Bayesian analysis is certainly the most complex of the techniques outlined here, and operates through the iterative definition of prior and posterior probabilities. By combining the probabilities of multiple environmental layers, a presence likelihood value is derived.