Skip to main content
Fig. 4 | BMC Medical Research Methodology

Fig. 4

From: Meta-analysis of binary outcomes via generalized linear mixed models: a simulation study

Fig. 4

Bias of τ2 in the REM when pi2=0.4, θ=1, 0≤τ2≤1 and n=50,100,250,1000. Estimation methods are: pluses - unconditional generalized linear mixed-effects model with fixed study effects (FIM), crosses - unconditional generalized linear mixed-effects model with random study effects (RIM), circles - a conditional generalized linear mixed-effects model with exact likelihood (NCHGN), triangles - a conditional generalized linear mixed-effects model with approximate likelihood (ABNM), rhombs - DerSimonian and Laird method (DL) and reverse triangles - restricted maximum likelihood method (REML). Light grey line at 0.95

Back to article page