Figure 4.

Bias, coverage probability (CP) and relative efficiency (RE) of estimates for δ = β ₁−β ₂ from restricted maximum likelihood (REML), Jackson's method (Jackson) and marginal method of moments (MMoM) based on 1000 simulations with data generated from bivariate random‐effects meta‐analysis model. The between‐study/within‐study variation ratio is close to 2 (i.e. between‐study variations ${\tau }_1^2={\tau }_2^2=0.5$ and median within‐study variation is 0.26). Within‐study correlations ρ _w are set to (−0.8,−0.5,0,0.5,0.8). Between‐study correlations ρ _b are set to (−0.8,−0.6,−0.4,0,0.4,0.6,0.8). Number of studies is m = 25. There are 30% missing data for each of the two outcomes.