The relative importance of each variable in predicting parasite occurrence in the Sevilleta host-parasite network.

Variable importance is measured as the reduction in predictive power by randomizing each variable, and the resulting variable importance scores are z-scores. Negative scores correspond to the proportional reduction in model performance as a result of variable randomization. Traits are ordered by importance to the predictive model, with the key predictive covariates in the upper left (e.g., litter size).