Extended Data Fig. 4: The effects of global change drivers and subsequent subcategories on disease responses with Log Response Ratio instead of Hedge’s g.

Here, Log Response Ratio shows similar trends to that of Hedge’s g presented in the main text. The displayed points represent the mean predicted values (with 95% confidence intervals) from a meta-analytical model with separate random intercepts for study. Points that do not share letters are significantly different from one another (p < 0.05) based on a two-sided Tukey’s posthoc multiple comparison test with adjustment for multiple comparisons. See Table S3 for pairwise comparison results. Effects of the five common global change drivers (A) have the same directionality, similar magnitude, and significance as those presented in Fig. 2. Global change driver effects are significant when confidence intervals do not overlap with zero and explicitly tested with two-tailed t-test (indicated by asterisks; t_80.62 = 2.16, p = 0.034 for CP; t_71.42 = 2.10, p = 0.039 for CC; t_131.79 = −3.52, p < 0.001 for HLC; t_61.9 = 2.10, p = 0.040 for IS). The subcategories (B) also show similar patterns as those presented in Fig. 3. Subcategories are significant when confidence intervals do not overlap with zero and were explicitly tested with two-tailed one sample t-test (t_30.52 = 2.17, p = 0.038 for CO₂; t_40.03 = 4.64, p < 0.001 for Enemy Release; t_47.45 = 2.18, p = 0.034 for Mean Temperature; t_110.81 = −4.05, p < 0.001 for Urbanization); all other subcategories have p > 0.20. Note that effect size and study numbers are lower here than in Figs. 3 and 4, because log response ratios cannot be calculated for studies that provide coefficients (e.g., odds ratio) rather than raw data; as such, all observations within BC did not have associated RR values. Despite strong differences in sample size, patterns are consistent across effect sizes, and therefore, we can be confident that the results presented in the main text are not biased because of effect size selection.