Skip to main content
The graph isn’t just a drawing. The Relationship analysis tool in the toolbar reads the structure and tells you, for a specific cause-and-effect question, which variables to watch together and which to keep out of your filters.

​
Running an analysis

  1. Click the Relationship analysis icon in the toolbar.
  2. Pick a Cause (X) and an Effect (Y) from the dropdowns. The chosen nodes get X and Y badges on the canvas.
The panel summarizes what sits between and around them:
  • Mediators variables the effect passes through on the way from X to Y.
  • Must watch together variables to control for; ignoring them biases the X→Y estimate.
  • Do-not-filter traps variables you must not condition on; filtering by them manufactures a spurious relationship.

​
Find hidden relationships

Find hidden relationships checks the structure for spurious paths between X and Y:
  • Spurious paths backdoor routes that make X and Y look related without causation. None found means the cause-effect pair is clean.
  • Variables to watch what to adjust for if a path exists.

​
Experiment simulation

Simulate experiment toggles an intervention view. X is marked as the treatment (T), showing the graph as if you forced X to a value. This helps you think through what an A/B test on X would actually move. Use Reset in the panel to clear the X/Y selection and badges.