Fig 1.
Diagram summarizing the methodology of meta-partition.
Step 1 involves assessment of heterogeneity with the homogeneity test and the value of I2, step 2 involves selecting the best moderator according to Logworth value with p-value correction, and step 3 involves integration of effect sizes in final subsets.
Fig 2.
Summary tree of the meta-partition of the main traits influencing sensitivity to wetland habitat loss.
Within each partition, effect sizes are organized from left (smaller) to right (larger). The best moderator that explains the variability of each partition is highlighted with a grey arrow, and numbers in brackets mean (in the form of X/Y, e.g. 9/10) the number of times this partition is kept (that is, is similar than in the original tree) in the sensitivity analysis (X) in comparison with the number of times that a partition is replicated (Y). The complete partition trees are provided in the S1 Appendix.
Table 1.
Results of the heterogeneity of the possible combinations of levels of the moderator taxon.
The selected combination is marked in bold.
Table 2.
Results of the heterogeneity of the different candidate moderators to perform the first partition.
The selected moderator is marked in bold.