Skip to main content
. 2019 Jan 8;10(3):360–375. doi: 10.1002/jrsm.1331

Table 5.

Simulation study results[Link]

(n 0,n 1) i = 6 Studies i = 12 Studies
Pseudo IPD REML HK DL Pseudo IPD REML HK DL
Coverage probability (5, 5)  96.2%  95.6%  89.4%  95.4%  95.2%  90.4%
(10, 10)  96.0%  95.6%  92.6%  95.4%  95.0%  92.4%
(40, 40)  94.8%  94.8%  87.8%  94.8%  94.6%  92.6%
Bias (5, 5)  0.0583  0.0508  0.0159 −0.0058 −0.0061 −0.0040
(10, 10)  0.0297  0.0300  0.0300 −0.0014 −0.0015 −0.0010
(40, 40) −0.0139 −0.0139 −0.0141 −0.0103 −0.0104 −0.0117
Mean squared error (MSE) (5, 5)  0.8581  0.8603  0.9958  0.4653  0.4668  0.5621
(10, 10)  0.6052  0.6050  0.6514  0.3062  0.3068  0.3521
(40, 40)  0.4050  0.4046  0.4034  0.2000  0.1998  0.1989

Abbreviations: DL, DerSimonian‐Laird; n 0, number of subjects in control group; n 1, number of subjects in exposure group; REML HK, restricted maximum likelihood with Hartung‐Knapp correction. Coverage of the calculated 95% confidence intervals, mean bias, and MSE obtained with the different methods.