This course is closed, registration is no longer possible.
This course teaches you to conduct meta-analyses according to contemporary best practices, using only free open-source software. It covers the basics of meta-analysis, including statistical models for meta-analysis, calculating, pooling, and converting effect sizes, visualizing the distribution of observed effect sizes using forest plots, and estimating and interpreting summary effect sizes. Particular attention is devoted to quantifying and explaining heterogeneity in effect sizes, a common challenge in applied meta-analyses. What sets this workshop apart are the advanced topics that address prevalent challenges in meta-analysis. We cover methods for testing and mitigating publication bias.
Furthermore, the course introduces the concept of three-level meta-analysis, a powerful tool for dealing with dependent data, which occurs when coding multiple effect sizes from the same paper or sample. In addition, the course introduces cutting-edge machine learning methods for exploring heterogeneity in effect sizes, including random forest meta-regression and Bayesian regularized meta-regression.
Finally, the course addresses computational reproducibility and open science in meta-analysis. The workshop is designed to be accessible for researchers with a foundational understanding of multiple linear regression (undergraduate level). Statistical concepts are explained at a conceptual level, rather than a mathematical level, and applied examples and tutorials are used to build a working knowledge of the methods. Those with prior experience in meta-analysis might enjoy the focus on open source software and advanced techniques, including machine learning methods.
The course does not cover the systematic literature search. All analyses are performed in R, predominantly using the packages metafor, metaforest, and pema.
