Fig. 3
From: Recent advances in methodology for clinical trials in small populations: the InSPiRe project
Meta-analyses of few studies are particularly challenging. Here, effect estimates from two studies in pediatric transplantation [51] are shown along with 5 different combined estimates based on several common approaches: two Bayesian analyses with different prior specifications, a normal approximation that is usually appropriate for large sample sizes, and two small-sample adjustments based on a Student-t distribution. We systematically investigated the long-run properties of popular meta-analysis procedures with a focus on few small studies [37, 38]. Figure reproduced from [38] under CC BY-NC-ND License [52]