Comparison of models using log-transformed counts of the explanatory variables.

Comparison of models using log-transformed counts of the explanatory variables (LP abundance of Leadbeater’s Possum, HBT abundance of hollow-bearing trees) fitted individually in negative binomial models for both target species in all four datasets (GG Greater Glider; MBP Mountain Brushtail Possum); the deviance is the change in –2 log(likelihood); the non-linear effect is the difference between a cubic smoothing spline with 2 d.f. and a linear model on the log scale; p-values are approximate because they are based on asymptotic properties.