Your first meta-analysis in 10 minutes
This walkthrough takes you from a blank MetaProc workspace to a finished forest plot, a GRADE summary of findings, and an exported report you can re-run anywhere — in about ten minutes. You do not need to write any R.
You’ll need: just the web app. Open MetaProc and you’re ready. (Prefer to install? See Web vs desktop.)
1. Load the example data
Section titled “1. Load the example data”You don’t need your own dataset to follow along. On the Data tab, choose Load
bundled example (example_pairwise.csv) — a small binary-outcome dataset with event
counts and totals per arm.
MetaProc reads the file, shows rows / columns / missing-value boxes, and renders a preview table. It also runs a few quiet checks: duplicate study-ID detection and (if enabled) biological-range and consistency rules.
2. Choose a template
Section titled “2. Choose a template”Go to the Templates tab and pick Pairwise — binary (2×2). MetaProc auto-guesses which columns play which role (events and totals for each arm, plus the study label); correct any mapping it got wrong using the role pickers.
Leave the defaults to start: effect measure RR, a random-effects model with the REML τ² estimator and the Knapp–Hartung adjustment on. Then press Run.
Not sure which template? The Plan tab asks a few PICO-style questions and recommends a template, effect measure, and model — then “Apply & go” pre-selects it here.
3. Read the forest plot
Section titled “3. Read the forest plot”Open the Forest plot tab. You’ll see each study’s effect estimate and confidence interval, the pooled summary diamond, and (on by default) the prediction interval. Use the controls to sort by effect size or precision, or relabel the x-axis. Export the figure as PNG, PDF, or SVG.
Alongside it: value boxes for the pooled estimate and CI, I², and k; a heterogeneity panel (Q, I² with CI, τ², prediction interval); a per-study estimates table; and a plain-language interpretation of the result.
Crucially, open the R code panel: it shows the exact escalc() / rma() calls that
produced everything above. Copy it and you can reproduce this analysis in plain R.
4. Rate the certainty with GRADE
Section titled “4. Rate the certainty with GRADE”Go to Appraise → GRADE. Start from High (for an RCT body of evidence) or Low (observational), then apply the downgrade/upgrade factors. MetaProc computes the certainty rating and — using a baseline risk you enter — fills a Cochrane-style summary-of-findings row straight from your live analysis: relative and absolute effect, k, N, I², and the rating.
5. Export a report and a reproducibility bundle
Section titled “5. Export a report and a reproducibility bundle”On the Report tab, choose the sections you want — Methods, included-studies table, pooled result, forest plot, heterogeneity, the reproducible R code, and software citations — and render to HTML (self-contained and offline), PDF, or .Rmd source.
Then press Download reproducibility bundle. You get a .zip containing a runnable
analysis.R, the dataset.csv it reads, the .Rmd, sessionInfo(), an renv.lock
pinning exact package versions, and a README. It re-runs in a clean R session and
reproduces your pooled estimate.
That’s a complete analysis
Section titled “That’s a complete analysis”In a few minutes you imported data, ran a pooled model on the real engines, read the heterogeneity and prediction interval, rated certainty with GRADE, and exported a report plus a bundle anyone can re-run. From here:
- Try subgroup analysis or the Robustness & diagnostics card (leave-one-out, influence).
- Explore Network meta-analysis or the drag-to-build Pipeline.
- Skim the product manual for the full list of what MetaProc can (and cannot) do.