ABSTRACT
In meta-regression, conventional regression techniques are adapted to study-level data. Subgroup analysis is a special case of meta-regression using a dummy-coded or categorical predictor.
This chapter provides a refresher of the idea behind regression in general, and then describes how the (random-effects) model can be extended to a meta-regression model. Furthermore, ways to assess the fit and adequacy of a meta-regression model are covered.
The second part of the chapter discusses multiple meta-regression, and introduces the concept of interactions terms. Limitations and pitfalls of (multiple) meta-regression are explored, including issues like overfitting and multicollinearity.
Lastly, statistical approaches to assess the robustness of meta-regression models, such as permutation tests and multi-model inference, are covered. All contents are supplemented by hands-on examples using R.
