Flexible Models for Meta-Analysis
Complex machine learning models have great potential to improve meta-analyses. These relationships between how effective an intervention is and the conditions under which it was studied can be complex and interrelated. Mapping them often needs a data-driven approach that is efficient, particularly as meta-analyses tend to have relatively few data points. Methods that can flexibly do this while still addressing the inferential demands of research syntheses can help inform where interventions will be most effective.