Glossary of Meta-Analysis Terms

Quick reference for terminology used throughout AI4Meta and the systematic review literature.

A – F

Allocation concealment: Procedures to prevent foreknowledge of treatment assignment in an RCT.
Confidence interval (CI): A range of values within which the true effect size is expected to fall, typically at the 95% level.
Cohen's d: A standardized mean difference calculated as the difference between two group means divided by the pooled standard deviation.
Effect size: A quantitative measure of the magnitude of a phenomenon (e.g., OR, RR, MD, SMD).
Fixed-effect model: Assumes all studies share one true effect size; differences are due to sampling error only.
Forest plot: A graphical display of estimated results from multiple studies, including the pooled summary.
Funnel plot: A scatter plot of study effect sizes against precision, used to assess publication bias.

GRADE: Grading of Recommendations, Assessment, Development, and Evaluations — a framework for rating evidence certainty.
Hedges' g: A bias-corrected version of Cohen's d, adjusting for small sample sizes.
Heterogeneity: Variability in study results beyond what would be expected from sampling error alone.
I² (I-squared): The percentage of variability in effect estimates due to heterogeneity rather than chance. Ranges from 0% to 100%.
Inclusion criteria: Predefined characteristics that studies must have to be included in the review (e.g., PICO elements).
Inverse variance method: A meta-analysis method that weights studies by the inverse of their variance (more precise studies get more weight).

Mean difference (MD): The absolute difference in means between two groups, used when studies measure outcomes on the same scale.
Meta-analysis: A statistical technique for combining results from multiple studies to obtain an overall estimate.
Meta-regression: An extension of meta-analysis that explores the relationship between study characteristics and effect size.
Network meta-analysis (NMA): Extends pairwise meta-analysis to compare multiple interventions simultaneously using direct and indirect evidence.
Odds ratio (OR): The ratio of the odds of an event in the treatment group to the odds in the control group.
PICO: Framework for defining the review question: Population, Intervention, Comparison, Outcome.
PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analyses — a reporting guideline.
Publication bias: The tendency for studies with positive or significant results to be published more often than those with null results.

Random-effects model: Assumes the true effect varies across studies; accounts for both within-study and between-study variability.
REML (Restricted Maximum Likelihood): A method for estimating between-study variance (τ²) in random-effects models.
Risk of bias: Systematic errors in study design, conduct, or analysis that may distort results.
Risk ratio (RR): The ratio of the probability of an event in the treatment group to that in the control group.
Sensitivity analysis: Analyses that test the robustness of the meta-analysis results to different assumptions or study exclusions.
Standardized mean difference (SMD): The difference in means divided by the pooled SD, used when studies measure the same outcome on different scales.
Subgroup analysis: Separate meta-analyses for different subsets of studies based on a shared characteristic.
Systematic review: A structured, transparent process for identifying, evaluating, and synthesizing all relevant studies on a specific question.
τ² (tau-squared): The estimated between-study variance in a random-effects meta-analysis.
Trim-and-fill: A method to estimate missing studies due to publication bias and calculate an adjusted pooled effect.