Fig. 2
Oilseed rape yield and pollinator community structure forest plots. Mean correlation coefficient (r) for the relationship between oilseed rape yield and measures of pollinator community structure (error bars ± 95% credible intervals) for a mesocosm and b field-based studies. To test if the effect of pollinator community structure was responsible for changes in yield above that resulting from overall insect abundance the correlation between abundance and yield was included as a moderator in all models (P values relate to the QM test of moderators where this effect was tested). The exception for this was for models directly testing the effect of abundance. There was no significant effect of either male sterility (mesocosm studies) or varietal breeding type (hybrid, restored hybrid or conventional). Correlations are back transformed from Fishers z values and final sample size (n) follows removal of studies with high influence (Cook’s distance > 1). Scaled functional divergence is shown, although results are qualitatively identical for measures when excluding pollinator control plots (µ = 0.28, CI: 0.01, 0.51)
