|
Abstract:
|
Random-effects meta-analysis models are commonly applied in combining and synthesizing information from multiple independent studies in medical research. However, one of the challenges of the models is that the true effects are inconsistent across studies and the between-study variance is unknown. Bayesian approaches allow us to incorporate sources of parameter uncertainty and borrow strength from other studies to estimate summary effects in meta-analysis. We developed weighted Bayesian random-effects models to take the study heterogeneity into account in the meta-analysis. Weighted Bayesian random-effects meta-analysis models in two data features, two sampling algorithms, and three inference adaptations were evaluated. Simulation studies were used to select an optimal weight, an optimal prior and an appropriate sampling algorithm. We demonstrate the performance of the weighted Bayesian random-effects meta-analysis models through a series of simulation studies.
|
