In meta-analysis, what is the key distinction between fixed-effects and random-effects models?

• A. A type of logistic regression
• B. Statistical method combining results from multiple studies; fixed-effects assume a common true effect across studies; random-effects assume true effects vary across studies
• C. A method to impute missing data
• D. A single study meta

Answer: (B)

The key distinction in meta-analysis is how the true effect is treated across studies. In a fixed-effects model, there is one true effect size shared by all studies, and observed differences are due only to sampling error within each study. In a random-effects model, the true effect is allowed to differ across studies, modeling a distribution of effects and estimating both the mean effect and the between-study variance. This usually yields wider confidence intervals when there is heterogeneity and is preferable when studies differ in populations, settings, or methods. Use fixed-effects when the studies are effectively identical and you want to generalize only to the included studies; use random-effects when you expect real variation in effects and you want broader generalization beyond the included set. The other options don’t describe meta-analysis models: it’s not a type of regression, not a missing-data imputation method, and not a single-study meta.