The paper re-evaluates random-effects meta-analysis, emphasizing the importance of estimation, prediction, and hypothesis testing. It discusses the distinction between inference on the mean of the random-effects distribution and the whole distribution. The authors argue that current methods often fail to provide key results and explore distribution-free, classical, and Bayesian approaches. They conclude that Bayesian methods naturally allow for full uncertainty, especially for prediction, but face challenges like computational intensity and sensitivity to prior assumptions. A simple prediction interval is proposed for classical meta-analysis, and Bayesian methods are extended with an example on 'set shifting' ability in eating disorders. The paper highlights the importance of quantifying heterogeneity, estimating the mean effect, study-specific effects, prediction, and testing. It discusses classical and Bayesian approaches under normal and non-normal assumptions for random effects, emphasizing the flexibility of Bayesian methods. An example illustrates the application of Bayesian analysis to set shifting studies, showing the predictive distribution and credible intervals. The paper concludes that Bayesian methods offer a more comprehensive approach to random-effects meta-analysis, particularly in handling uncertainty and providing predictive intervals.The paper re-evaluates random-effects meta-analysis, emphasizing the importance of estimation, prediction, and hypothesis testing. It discusses the distinction between inference on the mean of the random-effects distribution and the whole distribution. The authors argue that current methods often fail to provide key results and explore distribution-free, classical, and Bayesian approaches. They conclude that Bayesian methods naturally allow for full uncertainty, especially for prediction, but face challenges like computational intensity and sensitivity to prior assumptions. A simple prediction interval is proposed for classical meta-analysis, and Bayesian methods are extended with an example on 'set shifting' ability in eating disorders. The paper highlights the importance of quantifying heterogeneity, estimating the mean effect, study-specific effects, prediction, and testing. It discusses classical and Bayesian approaches under normal and non-normal assumptions for random effects, emphasizing the flexibility of Bayesian methods. An example illustrates the application of Bayesian analysis to set shifting studies, showing the predictive distribution and credible intervals. The paper concludes that Bayesian methods offer a more comprehensive approach to random-effects meta-analysis, particularly in handling uncertainty and providing predictive intervals.