Fig. 5.

Orchard plots showing the distribution of the indicator of small-study effect (model slope β_{1[small − study]}) for each meta-analysis and meta-analytic aggregation of β_{1[small − study]} (pooled β_{1[small − study]}). (A) Pooled β_{1[small − study]} across different meta-analyses and different types of effect size, indicating the pattern of small-study effects. (B) Pooled β_{1[small − study]} for each type of effect size. Solid circles = β_{1[small − study]} estimates obtained from each meta-analysis; the size of each solid circle is proportional to its inverse standard error (i.e. precision). Open circles = pooled β_{1[small − study]}. Thick error bars = 95% confidence intervals (CI). Thin error bars = prediction intervals (PIs). See more details in the legend of Fig. 2. All panels were made using orchard_plot() function in orchaRd R package (version 2.0) [62]