Biodiversity recovery
All possible measures of biodiversity for which the included studies reported data were used to calculate RRs; these measures included (a) species, gender, taxon or family richness; and (b) indices of species abundance, diversity, similarity, and composition. Using biodiversity measures calculated for different taxonomic levels or by different formulas enabled us to screen for differences in responses to restoration at different levels of ecological complexity [9], [13]. Each extracted datum was assigned to a single organism type. Data were analyzed using categorical, random-effects models because the data were most likely to satisfy the assumptions of these models [12]; the categories in the model were organism types.
To evaluate possible pseudo-replication effects, we calculated the mean RR for each of the three largest categories: macroinvertebrates, aquatic invertebrates, and vascular plants, using only one randomly selected effect size from each study. These mean RRs were similar to the means obtained when all effect sizes from each study were included, and the bias-corrected 95% bootstrap confidence interval of the reduced dataset overlapped with that of the entire dataset (Table S2). Therefore we retained all the data in our meta-analysis, similar to Rey Benayas et al. [9] and Vilà et al. [13].
Biodiversity recovery
All possible measures of biodiversity for which the included studies reported data were used to calculate RRs; these measures included (a) species, gender, taxon or family richness; and (b) indices of species abundance, diversity, similarity, and composition. Using biodiversity measures calculated for different taxonomic levels or by different formulas enabled us to screen for differences in responses to restoration at different levels of ecological complexity [9], [13]. Each extracted datum was assigned to a single organism type. Data were analyzed using categorical, random-effects models because the data were most likely to satisfy the assumptions of these models [12]; the categories in the model were organism types.
To evaluate possible pseudo-replication effects, we calculated the mean RR for each of the three largest categories: macroinvertebrates, aquatic invertebrates, and vascular plants, using only one randomly selected effect size from each study. These mean RRs were similar to the means obtained when all effect sizes from each study were included, and the bias-corrected 95% bootstrap confidence interval of the reduced dataset overlapped with that of the entire dataset (Table S2). Therefore we retained all the data in our meta-analysis, similar to Rey Benayas et al. [9] and Vilà et al. [13].
การแปล กรุณารอสักครู่..