Meta-analysis
Statistically pooling many studies into one overall estimate.
What it means
Meta-analysis is the quantitative synthesis of results from multiple studies addressing the same question, combining their effect sizes into a single weighted estimate with greater precision than any one study. Studies are typically weighted by their inverse variance, and models can assume a single common effect (fixed-effect) or a distribution of true effects across studies (random-effects). Beyond a summary number, it characterizes heterogeneity between studies and probes for small-study and publication-bias artifacts. Its conclusions are only as sound as the included studies: pooling biased or low-quality work yields a precise but misleading answer, the 'garbage in, garbage out' problem.
Examples
Combining forty trials of a therapy to estimate its average effect and to test whether it varies by patient age.
A retailer pools a hundred small price tests run across its stores to estimate the real effect of a discount, rather than trusting whichever single store happened to show the biggest jump.
Reviewers combine dozens of studies on class size and find a modest average benefit, plus a telltale pattern suggesting the small, flattering studies were the ones that got published.
First described in Term coined by Gene Glass (1976); modern methods by Hedges, Hunter & Schmidt.