Bayesian meta-analysis now – lets do it

摘要

Meta-analysis is a statistical tool that allows the analysis of results from various scientific studies, which are often not performed in the same place or using the same method. The data used in meta-analysis may be proprietary or may be obtained from literature or various databases. Meta-analysis is a crucial part of many systematic reviews, although not all systematic reviews include a meta-analysis. Therefore, meta-analysis should not be equated with systematic reviews. It is not incorrect to say that meta-analysis synthesizes the results from several studies and yields a new set of results. In other words, meta-analysis itself can, under certain conditions, be considered a method of producing new data. There are two basic models of meta-analysis, the fixed-effects and random-effects model. If there is no statistical or methodological heterogeneity and if it is not necessary to generalize the conclusions, a fixed-effect model is applied. On the other hand, if a generalized conclusion is to be obtained and if results from more than five different studies are taken into account, a randomized-effects model is appropriated. While the fixed-effect model assumes the existence of a common fixed parameter in all studies, the random-effects model assumes that such a common fixed parameter does not exist. Therefore, the fixed-effects meta-analysis model's total effect is an estimator of the combined effect of all studies. In contrast, the random-effect meta-analysis's full effect is an estimator of the mean value of the true effect distribution.