Meta-Analysis

Publication-ready forest plots. Rigorous statistics. Zero lines of R code.

Meta-analysis is the quantitative heart of a systematic review — the step where individual study results become pooled estimates. It's also where most researchers hit a wall: learning R, debugging metafor scripts, formatting plots for journal requirements. mapped's R-based statistical engine handles the computation while giving you full control over model specification and interpretation.

---

Statistical Models

mapped supports the full range of meta-analytic models:

Fixed-Effects Model

Assumes all studies estimate the same underlying effect. Appropriate when:

• Studies share similar populations, interventions, and designs
• Heterogeneity is expected to be minimal
• Weighted by inverse variance (Mantel-Haenszel or Peto methods also available)

Random-Effects Model

Assumes true effects vary across studies. Appropriate when:

• Studies differ in populations, settings, or intervention details
• Heterogeneity is expected and meaningful
• DerSimonian-Laird, REML, or maximum-likelihood estimators available
• Knapp-Hartung adjustment for confidence intervals

Mixed-Effects Model

Combines fixed moderator effects with random study-level effects. Used for meta-regression and subgroup analysis with continuous or categorical moderators.

---

Effect Measures

mapped supports all standard effect measures:

Select the appropriate measure for your data, and mapped handles the transformation, weighting, and pooling.

---

Forest Plots

The signature visualization of any meta-analysis. mapped generates publication-ready forest plots that include:

• Individual study effect estimates with 95% confidence intervals
• Study weights (proportional to box size)
• Pooled effect estimate (diamond)
• Heterogeneity statistics (I², τ², Q statistic, p-value)
• Prediction interval (optional, for random-effects models)
• Subgroup subtotals (when subgroup analysis is performed)

Plots are fully customizable:

• Font size, colors, and layout
• Study ordering (by year, effect size, weight, or custom)
• Label formatting and axis ranges
• Export to PNG, SVG, PDF, or EPS for journal submission

---

Funnel Plots

Funnel plots visualize the relationship between study effect sizes and precision, helping detect publication bias:

• Standard funnel plot — effect vs. standard error
• Contour-enhanced funnel plot — overlays significance regions
• Trim-and-fill funnel plot — shows imputed missing studies

Asymmetry in the funnel suggests potential publication bias — a critical issue that must be addressed in your manuscript.

---

Publication Bias Tests

Beyond visual inspection, mapped runs formal statistical tests:

Results are automatically reported with the interpretation needed for your manuscript.

---

Subgroup Analysis

Test whether the effect differs across predefined subgroups:

• Categorical moderators — study design, geographic region, risk of bias level, intervention type
• Between-group heterogeneity test — quantifies whether subgroup differences are statistically significant
• Subgroup forest plots — separate pooled estimates per group with a test for interaction

mapped ensures subgroup analyses are pre-specified (to avoid data dredging) and correctly interpreted.

---

Meta-Regression

For continuous moderators or more complex relationships:

• Single-predictor models — does year of publication or sample size predict effect?
• Multi-predictor models — multiple moderators simultaneously
• Bubble plots — visualize the relationship between a moderator and effect size
• R² analog — proportion of heterogeneity explained by the moderator

---

Sensitivity Analysis

Test the robustness of your findings:

• Leave-one-out analysis — re-run the meta-analysis excluding each study in turn
• Influence diagnostics — identify studies with disproportionate impact on the pooled estimate
• Cumulative meta-analysis — see how the pooled estimate changes as studies are added chronologically
• Exclusion by quality — compare results when restricting to high-quality studies only

---

Heterogeneity Assessment

mapped reports comprehensive heterogeneity statistics:

• Q statistic — tests the null hypothesis that all studies share one true effect
• I² statistic — percentage of variability due to true differences rather than chance
• τ² (tau-squared) — absolute measure of between-study variance
• Prediction interval — range where the true effect of a future study would likely fall

Interpretation guidelines are displayed alongside the statistics — no need to look up thresholds.

---

Export and Reporting

All outputs are designed for direct manuscript insertion:

• Plots: PNG, SVG, PDF, EPS
• Tables: Word, Excel, CSV, LaTeX
• Statistical reports: formatted text blocks ready for Results sections
• Raw data: R script equivalent if you want to verify or extend the analysis independently

---

Why This Step Matters

A meta-analysis is only as credible as its methodology. Incomplete heterogeneity assessment, missing publication bias tests, or poorly specified models are the most common reasons for major revisions. mapped runs all the standard tests, visualizations, and sensitivity analyses that peer reviewers expect — in a guided workflow that doesn't require statistical programming expertise.

---

Next step: With your analysis complete, learn about Collaboration → to manage your team throughout the process.