Advanced Research Methods: Meta-Analysis

Vocabulary practice flashcards covering foundational concepts, study designs, statistical models, weighting methods, and bias detection in meta-analysis.

Meta-analysis

A computational method for combining and synthesizing results across multiple empirical studies to evaluate an overarching research question, typically asking whether there is an effect of $x$ on $y$.

PICOS framework

A systematic approach used to define study inclusion and exclusion criteria, standing for Participants, Intervention, Comparison, Outcomes, and Study type.

Grey literature

Non-commercially published research materials—such as PhD theses, dissertations, and conference proceedings—searched during a systematic review to mitigate publication bias.

PRISMA flow diagram

A standardized flow diagram used to transparently document the study selection process in a systematic review, tracing records from identification and screening to eligibility and final inclusion.

Standardised effect size

A metric that quantifies the magnitude of an experimental effect or relationship independently of the original scale of measurement, allowing comparisons across different studies.

Cohen's $d$ / Hedges' $g$

Types of standardized effect sizes used to quantify the magnitude of differences between two group means.

Pearson's $r$

A standardized effect size metric extracted in meta-analyses to measure the strength and direction of linear associations between continuous variables.

Odds ratio

An effect size metric commonly used to represent and compare probabilities of events for dichotomous outcome data.

95% Confidence Interval (in meta-analysis)

The range of values expected to contain the true estimate 95 out of 100 times upon repeated replication; calculated under normality assumptions as $\text{Effect size} ± 1.96 \times SE$.

Inverse variance weighting

A method of weighting studies in a meta-analysis where each study's weight equals the reciprocal of its squared standard error ($\frac{1}{SE^2}$), giving greater influence to more precise studies.

Forest plot

A standardized graphical representation of meta-analytic data, displaying each individual study's effect size as a square, its confidence interval as a horizontal line, and the pooled aggregate effect as a diamond.

Fixed effect model

A statistical model (also called a common-effect model) assuming a single identical true effect size underlies all included studies, treating any differences in observed outcomes entirely as sampling error.

Random effects model

A statistical model assuming that true effect sizes vary across studies due to differences in participants, interventions, or outcomes, aiming to estimate the mean of this distribution of underlying effect sizes.

Statistical tests for heterogeneity

Statistical tests and metrics—such as Cochran's $Q$, $\tau^2$, and $I^2$—used to quantify whether differences across study effect sizes exceed what would be expected by sampling error alone.

metafor

An R package used to perform meta-analyses, which includes functions such as rma() for model fitting and forest() for plotting results.

File drawer problem

A form of publication bias where studies with non-significant or null findings are withheld from publication and left in filing drawers, creating an unrepresentative published literature.

Garbage in, garbage out (GIGO)

The concept that if the input studies synthesized in a meta-analysis are methodologically flawed, unrepresentative, or biased, the resulting pooled aggregate conclusions will also be invalid.

Funnel plot

A scatterplot displaying individual study effect sizes against their precision (or standard error), used to visually evaluate whether studies form a symmetrical inverted funnel or show signs of publication bias.

Funnel plot asymmetry

A visual skewness in a funnel plot—particularly an absence of smaller studies with null or weak effects at the bottom—which often suggests the presence of publication bias.

Cochrane Collaboration

An international organization producing systematic reviews and meta-analyses of healthcare interventions to synthesize high-quality medical evidence and improve clinical decision-making.