Concept
meta-analysis
A method that statistically combines the effect sizes of similar studies to obtain a common estimate and the reasons for differences; sensitive to the file-drawer problem and measurement error.
Meta-analysis statistically combines the effect sizes of many studies asking similar questions to obtain a common estimate and to look for the reasons for differences (moderators). Gene Glass introduced the term in 1976. The steps: write the question and search rules, collect the studies (including unpublished ones), extract an effect size from each, combine them, explain the differences and test for bias.
It has two traps: the file-drawer problem that Rosenthal discussed (studies that do not turn out significant go unpublished) and the fact that combining does not repair poor measurement. Example: a 2023 meta-analysis of cultivation and social media combined 460 effect sizes from 66 independent samples and found an overall effect of .123; a 1997 cultivation meta-analysis had found an average effect of .09. See effect size, replication crisis.
Example
Combining 460 effect sizes from 66 samples on social media use and perceptions of the world.