Although including all protected areas revealed the same trend
in the model-averaged, standardized coefficients, the inclusion of
the positive-health reserves weakened the relationship as expected
(confidence bounds overlapping zero; Table S4; Fig. S2). This confirms
the problem of including the protected areas where health
was increasing (positive index). The inclusion of Category III protected
areas in the ‘high-protection’ reserves made no qualitative difference to the model ranking (Table S5) or model-averaged,
standardized coefficient rankings (Fig. S3). Treating predictors as
proportional gave qualitatively similar results (Tables S6–S11;
Fig. S4).