Research the Predictor Variables

The next step is to research the phenomenon being modeled and specifically, what predictor variables might influence the response variable.

Define the potential response variables

Predictor variables for a plant-based response (height, potential habitat) might include minimum temperatures, maximum temperatures, water availability, absolute aspect, and soil type. Predictor variables for a predator might be similar to plants but could also include prey species, parasites, and cover.

Pre-determine the shape of the response

The modeling method should partly be selected based on the expected shape of the response curve. The most powerful statistics tools are based on linear regression but there are few responses that are linear for any large extent.

Habitat models should go to zero outside the viable area for the species.