Figure 2. Variable importance of the predictors of regional
alien species richness of plants (A, C, and E; n = 347 regions)
and birds (B, D, and F; n = 423 regions) using three statistical
techniques: hierarchical partitioning (A and B), random forests (C and
D), and mixed linear models (E and F) (results for the other taxonomic
groups are shown in Figs. S4–S7). For hierarchical partitioning, the
individual (green) and shared (yellow) contributions are shown. The
conditional (threshold = 0.95) variable importance of all predictors of
random forests is shown in light blue bars. Mixed models refer to the
beta weights (brown circles) and their 95% confidence interval. For the
HP and LMM, region area, and the relative and absolute measures of
economy and population were log-transformed, and precipitation was
square-root transformed. GDPc: Gross Domestic Product per capita, and
HPD: Human Population Density.