Analyses:
To compare nekton recruitment to oyster reefs versus unstructured control habitat, we calculated log response ratios (LRRs, Hedges et al. 1999) with 95% confidence intervals by family (Appendix S2 of Davenport et al. 2021). An LRR with CIs > 0 implies that nekton recruitment was enhanced by oyster reefs, an LRR with CIs < 0 implies the opposite, and an LRR with CIs that include 0 implies no difference in recruitment between oyster reef and control habitats. For each research question, we assessed data publication bias with funnel plots (Appendix S3 of Davenport et al. 2021) and data availability with mosaic and violin plots (Appendix S4 of Davenport et al. 2021).

To assess whether recruitment enhancement varied as a function of reef type (natural or restored) and nekton family, we conducted linear mixed model analyses separately for intertidal and subtidal reefs that included two categorical factors (reef type and nekton family) and their interaction as fixed effects, with study as a random effect. We conducted mixed model analyses separately by tidal zone since subtidal and intertidal habitats harbor different nekton communities (Lehnert and Allen 2002), and subtidal and intertidal oyster reefs are distributed unevenly in our database: intertidal reefs are more prevalent on the Atlantic coast and subtidal reefs are more common in the Gulf of Mexico (Fig. 1; Appendix S1: Table S1 of Davenport et al. 2021). Furthermore, the reef characteristics in our database differed across tidal zones (Appendix S4: Fig. S1-S12 of Davenport et al. 2021). We removed any family that was represented by fewer than three independent studies and/or fewer than 10 independent LRRs ("NA" on Fig. 2 of Davenport et al. 2021). Results of randomization and resampling in support of these criteria are presented in Appendix S5 of Davenport et al. (2021).

To evaluate the effects of restored reef characteristics (i.e., reef size, vertical relief, and age) on recruitment enhancement, we focused on restored reefs only, and performed linear mixed effect regression models with orthogonal polynomials (first and second order) for the continuous reef characteristics as fixed effects, and study as a random effect. We conducted separate analyses for each reef characteristic, nekton family, and tidal zone. We did not model families represented by fewer than three independent studies and/or 10 independent LRRs covering different values of the reef characteristics (e.g., reef sizes, vertical reliefs, or ages; "insufficient data" in Figs. 3-8 of Davenport et al. 2021), and results are not reported or plotted for models that did not converge. Since not all families were represented at all values of each reef characteristic (e.g., all vertical reliefs or sizes), we performed separate analyses for each family and included all available data (but for combined-family approaches, see Appendix S2 (Davenport et al. 2021) for methods and Appendix S5 of Davenport et al. 2021 for results ). We also examined the influence of tidal elevation on recruitment enhancement for each family using linear mixed effect regression models with orthogonal polynomials (first and second order) for tidal elevation as a fixed effect and study as a random effect (Appendix S5 of Davenport et al. 2021). Where models indicated that predictor variables are on substantially different scales (i.e. reef size) we rescaled using natural log transformation.

All models were weighted by sample size (Appendix S2 of Davenport et al. 2021). All analyses were conducted in R 3.6.1 (R Core Team, 2019) on the RStudio IDE 1.2.1335 (RStudio Team, 2019).

BCO-DMO processing description:
- Adjusted field/parameter names to comply with BCO-DMO naming conventions;
- Missing data identifier ‘NA’ replaced with 'nd' (BCO-DMO's default missing data identifier);
- Added a conventional header with dataset name, PI names, version date;
- Removed commas from the "embayment" column;
- Rounded oys_fish, con_fish, and LRR to 3 decimal places;
- Replaced "Humphries 2011" with "Humphries et al. 2011" in the study_name column.

meta-analysis_data.csv