Skip to main content

Table 1 Summary of Simulation Parametersa

From: A comparison of analytic approaches for individual patient data meta-analyses with binary outcomes

Parameters

Values

IPD-Meta-analyses generated:

M = 1000

(Number of studies, number of subjects per study, total average sample sizes)b:

(K, n i N) {(5,100,500), (15,33,500), (15,200,3000), (5,357,500), (15,98,500), (15,588,3000)}

Fixed effects (intercepts):

β 0 = − 0.85

Prevalence of the outcome

π = 30%

Fixed effects (Slopes):

β 1 = 0.18

Random effects distribution:

Normal

Random effects variances:

{τ 20 , τ 21 }  (0.05, 1, 4)

Correlation between random effects:

ρ  (0,0.5)

  1. aIn a sensitivity analysis, we extended the number of studies to 50 with an average sample size of 9000 and reduced the prevalence of the outcome to 5%. The prevalence of the outcome was fixed to 30% by setting the value of the intercept β 0 to –0.85
  2. bThe number of subjects per study was reported for only large studies when data sets were generated with imbalanced study sizes (bold text: 25% large studies-10 times more subjects)