Skip to main content
. Author manuscript; available in PMC: 2015 Dec 1.
Published in final edited form as: Stat Methods Med Res. 2011 Dec 4;24(6):836–855. doi: 10.1177/0962280211430889

Table 1.

Estimates of the bias, true standard error (SD), model based standard error (SE), coverage probability (CP) and relative efficiency (RE) of common odds ratio estimated from all studies (in log scale) in 5000 simulations based on Sarmanov beta-binomial model and independent beta-binomial model, with different number of studies n, for different within study correlations ρ. (a1, b1, a2, b2) = (0.5, 0.5, 0.5, 0.5).

Sarmanov model Independent model
n ρ Bias SD SE CP (%) RE Bias SD SE CP (%) RE
20 0 −0.005 0.439 0.415 92.6 1.000 −0.005 0.438 0.416 93.4 0.995
0.2 0.003 0.388 0.378 94.4 1.000 0.003 0.395 0.416 96.0 0.826
0.4 0.003 0.321 0.336 96.8 1.000 0.004 0.352 0.416 97.7 0.652

40 0 −0.001 0.302 0.294 93.8 1.000 −0.001 0.302 0.295 94.4 0.993
0.2 0.005 0.271 0.268 94.5 1.000 −0.006 0.272 0.295 96.7 0.825
0.4 0.002 0.228 0.233 96.1 1.000 0.002 0.247 0.295 98.2 0.624

60 0 0.000 0.250 0.240 93.7 1.000 0.000 0.250 0.241 94.0 0.992
0.2 0.004 0.222 0.219 94.4 1.000 0.004 0.222 0.241 96.7 0.826
0.4 0.005 0.185 0.188 95.5 1.000 0.006 0.201 0.241 98.1 0.609