Term | Definition (with associated parameters, if any) |
---|---|
Effect size | A measurement of effect (usually state of a single group, comparison between groups, or association, see Table 2). In a meta-analytic model, it becomes the response variable (noted as zi in the formulas) |
Sampling variance | A measure of uncertainty in effect size (noted as vi). Its inverse is often called ‘weight’ (the square-root of weight is ‘precision’, and the square root of sampling variance is ‘sampling standard error’) |
Meta-analysis | A statistical method to aggregate effect sizes from studies on the same or similar topics, by assigning different weights based on sampling variance of effect sizes. Strictly speaking, in a formal (weighted) meta-analysis, sampling variance needs to be incorporated and it is assumed to be known (Table 2) |
Overall mean (effect) | An average effect size based on a meta-analytic model (noted as \({\beta }_{0}\) and its standard errors, se(\({\beta }_{0}\))) |
Heterogeneity | An indicator of consistency among effect sizes, or an extent of variation around the overall effect (\({\beta }_{0}\)); heterogeneity can be quantified by absolute measures, such as \({\tau }^{2}\), or relative measures, such as I2 |
Meta-regression | A regression model which extends a meta-analytic model with a moderator(s), aiming to explain heterogeneity (quantified as R2) and quantifying the effect of a moderator (noted as, for example, \({\beta }_{1}\)) |
Publication bias tests | A set of statistical methodologies to detect and correct for publication bias, where a subset of results (positive findings) is more likely to be published and present in the meta-analytic dataset than otherwise |
Sensitivity analysis | A set of statistical analyses that checks the robustness of one’s main analysis; if sensitivity analysis shows different results (qualitatively and/or quantitively), then we must doubt the robustness of the main findings |