Insyt
All charts in the library

Compare and rank

Forest plot

Compares study or subgroup effects and uncertainty, optionally including a pooled effect.

  • Plot
  • Effect comparison
  • Uncertainty

What it answers

How do effect estimates and their uncertainty compare?

Shape
Effect dots with horizontal intervals
What it shows
Effect direction, precision and heterogeneity.

Example

The question

Did the loyalty offer lift spending in every customer group?

Effect of the loyalty offer on monthly spend by group, with 95% intervals (EUR)

Effect of the loyalty offer on monthly spend by group, with 95% intervals (EUR)Squares show each group’s estimated effect with 95% intervals; the pink diamond is the overall effect (+6.5 euros, 2 to 11). Intervals for 50 to 64 and 65 and over cross zero, so the lift there is uncertain.Under 3030 to 4950 to 6465 and overNewOverall-10-505101520Change in monthly spend (EUR)+6+9+4-1+12+6.5
Effect of the loyalty offer on monthly spend by group, with 95% intervals (EUR)Squares show each group’s estimated effect with 95% intervals; the pink diamond is the overall effect (+6.5 euros, 2 to 11). Intervals for 50 to 64 and 65 and over cross zero, so the lift there is uncertain.Under 3030 to 4950 to 6465 and overNewOverall-10-505101520Change in monthly spend (EUR)+6+9+4-1+12+6.5
Squares show each group’s estimated effect with 95% intervals; the pink diamond is the overall effect (+6.5 euros, 2 to 11). Intervals for 50 to 64 and 65 and over cross zero, so the lift there is uncertain.

Illustrative data. It belongs to no company.

When to use it

Use it when

Comparing studies, subgroups or model effects.

Avoid it when

Mixing effect scales or assuming all estimates are independent.

How to read it

Read each effect dot, its interval and the no-effect reference.

Note

Define effect scale and no-effect reference.

The data you need

Layout
Wide interval summary
Fields
Study or Term: text; Estimate,Lower,Upper: numbers
One row is
One estimate and its lower and upper interval bounds. What is the grain?
Rules
Keep a common effect scale and specify the no-effect reference. Log-scale ratio effects require positive bounds.
Optional
Weight,Group,PooledIndicator,ModelName.
Preparation
Fit or obtain a suitable model/estimate first, then pair estimates with correctly labelled bounds or standard errors. Preserve coverage and scale information.
Units
Estimates and interval endpoints share an effect/measurement unit unless transformed. Coverage such as 95% is a model-based probability statement, not a data-share denominator.
Example of the input layout: Wide interval summaryA wide table: one row per item, with a separate column for each measure or period.Study or TeEstimateLowerUpper…ABC
A wide table: one row per item, with a separate column for each measure or period.

Worked example

StudyA effect 0.20, bounds 0.05-0.35; B0.10,-0.08-0.28. OnlyA's supplied interval excludeszero; no pooled effect is implied.

StudyEstimateLowerUpper
A0.200.050.35
B0.10-0.080.28
All charts in the library