← allmeta Paper Studio →

Forest Plot Viewer

Paste study-level effect sizes (already on the analysis scale — e.g. logOR, logRR, MD, SMD). Fixed-effect and Paule-Mandel random-effects pooling, live SVG render, SVG / PNG / JSON export.

What to do on this page — pool the effects and check the model

Combine the per-study effects into one pooled estimate and inspect the heterogeneity.

  1. Your checked data arrives from Extraction via the shared workspace (or paste / load it here).
  2. Choose fixed-effect or random-effects pooling.
  3. Read the pooled estimate, its 95% CI, and heterogeneity (I², τ²); export the figure if needed.
  4. Click Paper Studio → (top right) to write it up.

The tool does

Computes the pooling maths and renders the forest plot (results are reproducible / verifiable).

You must check

That pooling is appropriate — similar-enough studies, enough of them — and choose the model.

Studies

Data
Display

Both models are shown: fixed-effect inverse variance with a normal CI, and Paule–Mandel random effects with modified Hartung–Knapp (variance floor 1), matching metafor test="adhoc". I² uses τ² and the typical sampling variance; the historical Q-based I² is also retained in JSON. Prediction intervals are approximations and can be unstable with few studies. Numerical table entries remain on the input scale; exponentiation changes axis labels only.

Preview

Leave-one-out sensitivity · CMA-style

Click a study to drop it and watch the pooled estimate update live; the forest shows the random-effects pool recomputed with each single study omitted (metafor's leave1out). Drops here are a what-if sensitivity view — they do not change your data, exports, or the bus.

Pooling: fixed-effect (inverse-variance) and random-effects (Paule-Mandel τ² + Hartung-Knapp-Sidik-Jonkman SE correction with floor q* ≥ 1, per Cochrane Handbook §10.10.4). The dashed band is the 95% prediction interval (μ ± tk−1 × √(SEHKSJ2 + τ2) per Cochrane Handbook v6.5, matches metafor::predict; undefined for k<2). Computed client-side; no data leaves the browser.