A forest plot is the standard figure for a meta-analysis: it shows every study’s effect estimate and confidence interval on one row, sizes each marker by the study’s weight, and summarises them with a pooled diamond. This forest plot generator takes your studies on a ratio or difference scale, runs the inverse-variance pooling, and draws the plot with a null reference line you can download for your manuscript.
Forest plot generator
Paste your studies, choose a fixed-effect or random-effects model, and get a publication-style forest plot with a pooled diamond, ready to download as an SVG.
Drag and drop or click. CSV, TSV, Excel; header row with columns like study, label, author, trial, estimate.
5 studies plotted. Marker area is proportional to study weight; the PNG renders at three times figure size for print.
What the forest plot shows at a glance
A forest plot compresses an entire meta-analysis into one readable figure. Reading top to bottom, each row is a study; reading left to right, the marker sits at the study’s point estimate and the line through it is the 95% confidence interval. The eye is drawn first to two things: how far each marker sits from the line of no effect, and how wide the intervals are. A tight interval near the null says a study found little and found it precisely; a wide interval says the study was small or noisy, which is exactly why its marker is drawn smaller. Our guide to interpreting a forest plot walks through each of these signals in turn.
Why the markers are different sizes
The square for each study is scaled to its inverse-variance weight, so precise studies dominate the picture and the summary. That weighting is the same machinery behind the pooled estimate, and whether you add the between-study variance to it depends on your model. The choice between a fixed-effect and a random-effects model changes how evenly the weights are spread and how wide the diamond becomes, which is why this generator lets you switch between them and redraw instantly.
A worked example and the traps to avoid
Suppose five trials report risk ratios between 1.05 and 1.66, most with intervals that cross 1. Pooled under a random-effects model the diamond might land at a risk ratio of about 1.30 with an interval of 1.10 to 1.53, so the summary clears the null even though three of the five individual studies did not. That is the point of pooling: precision accumulates. The diamond, not any single row, carries the answer, and its width tells you how confident that answer is. The plot is drawn on a logarithmic axis for ratio measures so that a halving and a doubling sit the same distance either side of the null.
Common mistakes researchers make
The first trap is reading the forest plot without reading the heterogeneity beside it. A diamond drawn over markers pointing in different directions is hiding real disagreement, and you should quantify it with the I-squared and tau-squared statistics before you trust the pooled value. The second is mixing effect measures on one plot, such as pasting odds ratios next to risk ratios; put every study on a single scale first, a step our guide to running a meta-analysis sets out. The third is treating a wide diamond that crosses the null as proof of no effect when it often just means too few studies. If the synthesis behind the figure needs to stand up to peer review, our meta-analysis service builds the model, the plot, and the write-up as one deliverable.
How it works
Ratio measures are pooled on the log scale; the plot exponentiates the axis labels back. For each study i with estimate yi and standard error si:
fixed-effect weight: wi = 1 / si^2
random-effects weight: wi = 1 / (si^2 + tau^2)
pooled estimate = sum(wi * yi) / sum(wi)
pooled SE = sqrt( 1 / sum(wi) )
diamond half-width = 1.96 * pooled SE (on the analysis scale)
The marker area is proportional to wi, the horizontal line is the study confidence interval, and the dashed vertical line marks no effect (1 for ratios, 0 for differences). The between-study variance tau-squared uses the DerSimonian and Laird estimator.
Frequently asked questions
- How do you prepare a forest plot?
- Start from the effect estimate and 95% confidence interval (or standard error) for every included study, all on one scale, and decide whether you are pooling a ratio measure such as a risk ratio or a difference such as a mean difference. Enter one study per line, choose a fixed-effect or random-effects model, and the generator plots each study as a marker with its interval and adds a pooled diamond. The only preparation that matters is getting every study onto the same effect measure first.
- What should a forest plot look like?
- Each study sits on its own row as a marker at its point estimate with a horizontal line spanning the confidence interval, the marker sized by the study's weight in the pool. A vertical reference line marks no effect, which is 1 for ratio measures and 0 for differences. The pooled result appears at the foot as a diamond whose width is the confidence interval of the summary estimate, so a reader can see the overall answer and its precision in a single glance.
- How is the weight in a forest plot calculated?
- Each study is weighted by the inverse of its variance, so a study with a smaller standard error gets more weight and a larger marker. Under a fixed-effect model the weight is one divided by the study's squared standard error; under a random-effects model the between-study variance, tau-squared, is added to that squared standard error before inverting, which pulls the weights closer together. The generator reports each study's percentage weight next to its interval.
- Can I create a forest plot in Excel?
- You can force a forest plot out of a stacked bar or error-bar chart in a spreadsheet, but it is slow and error-prone because the weighting, the log scaling of ratio measures, and the pooled diamond all have to be built by hand. This generator does the inverse-variance pooling and the scaling for you and exports a clean SVG, which scales without blurring in a manuscript and is far easier to update when a study is added or a number changes.
- What are the key elements of a forest plot?
- Five things: the study labels down the left, a marker for each study's point estimate scaled to its weight, a horizontal line for each confidence interval, a vertical line at the value of no effect, and a pooled diamond summarising the meta-analysis. Many journals also expect the numeric estimate and interval printed beside each row and a clear axis label stating the effect measure and, for ratios, that the axis is logarithmic.
- How do you make a forest plot step by step?
- First put every study on the same effect scale and gather its estimate with a confidence interval or standard error. Second, choose a ratio or difference measure so the tool knows whether to work on the log scale. Third, paste the studies one per line and pick a random-effects or fixed-effect model. Fourth, read the markers, intervals, and the pooled diamond, and check the heterogeneity before you trust the summary. Fifth, download the SVG for your manuscript.
The guides behind this tool
More review and meta-analysis tools
Meta-analysisPool study effect estimates with fixed-effect and random-effects models, with the heterogeneity statistics shown.Open tool Heterogeneity (I-squared)Quantify how much your studies disagree: Cochran's Q, I-squared, H, and the between-study variance tau-squared.Open tool Odds ratio / risk ratioTurn a 2x2 table of events into an odds ratio, risk ratio, and risk difference, each with a 95% confidence interval.Open tool