Meta-analytic methods in educational research : issues and their solutions

Meta-analysis is a valuable statistical technique for synthesising the available educational research literature on a particular topic, owing to its comprehensive and relatively unbiased approach to analysis. Since its conception by Glass in 1976, meta-analysis has been embraced by many researchers in education, psychology, and other disciplines, leading to the evolution of several very distinct methods for conducting meta-analysis - each with their own unique implications. The present paper starts with a discussion of the purpose and value of meta-analysis. This is followed by an overview of the features of the most common meta-analytic techniques (fixed effects models and random effects models), and the emerging technique of multilevel modeling meta-analysis. Historical and contemporary issues in meta-analysis are detailed, with particular attention to publication bias, generalisability and the assumptions inherent in various meta-analytic techniques, multivariate analyses and the independence of effect sizes, and power. Suggestions for how to address these issues will be provided. Finally, guidelines for selecting appropriate meta-analytic methods will be presented, as well as suggested resources for researchers wishing to conduct a meta-analysis. The emphasis of the paper is on understanding the various approaches to conducting responsible meta-analysis and their implications, rather than providing a prescriptive account of performing the technique.