What a forest plot shows
A forest plot is a visual rendering of study-level effect estimates and uncertainty intervals, plus one or more pooled estimates. Coventra does not treat the plot as a separate analysis; it uses the selected model settings and extracted rows from the analysis run.
Each study marker represents the study effect. The horizontal line represents the confidence interval. Marker size is tied to model weight. Summary diamonds represent common-effect and/or random-effects pooled estimates depending on display settings.
Scale and back-transformation
Ratio measures are analyzed on the log scale and usually displayed after exponentiation. OR, RR, and HR have a null value of 1 on the display scale and 0 on the log scale.
Difference measures such as MD, SMD, RD, and raw mean have a null value of 0. Axis labels must make the direction of benefit clear because the same numeric estimate can imply benefit or harm depending on outcome direction.
- Back-transform is user-changeable in plot options.
- Log-spaced ticks for OR, RR, and HR are user-changeable.
- Left and right axis labels are user-changeable and should be set for journal figures.
- Study sorting can be changed by effect size, year, random weight, common weight, sample size, or label.
Default plot assumptions
- Default style follows standard forest plot conventions.
- Default resolution is 150 dpi, with 300 dpi available for print-oriented export.
- Common-effect and random-effects diamonds are shown by default.
- Prediction interval is hidden by default.
- Study labels are shown in the forest table; label wrapping and columns are user-changeable.
- Users can exclude studies from a plot view without deleting project data.
What the plot does not prove
A well-rendered forest plot does not prove that the effect-size formula, extraction values, study eligibility, or model choice was correct. It only displays the model result that was produced from the selected inputs.
Prediction intervals, heterogeneity labels, and study weights should be checked against the numerical output. Journal figures should be reviewed after any model-setting change, because the plot can update while manuscript text remains stale.
Reporting checklist
- State the effect measure and whether ratio measures were displayed on a log-scaled axis.
- State whether common-effect, random-effects, or both summaries were shown.
- State whether a prediction interval was shown and how it should be interpreted.
- State which studies were excluded from the plotted analysis, if any.
- Confirm that labels, direction of benefit, and outcome timepoint match the manuscript.