What is it?
A Scatter Diagram plots paired observations for two numerical variables. It helps show whether higher or lower values of one tend to occur alongside changes in the other.
What's it really for?
It is a useful defence against confident stories. Sometimes the suspected relationship appears clearly. Sometimes the points form a cloud and a cherished explanation has to return to the waiting room. Either result improves the investigation.
You know when…
Everyone knows overtime is causing defects. The two figures are quoted in the same meeting and nods occur in several parts of the room. When the paired observations are plotted, the expected diagonal line fails to appear. The conversation can finally move on.
Have a go
- State the relationship being explored. ☐ Gather paired observations from the same cases or periods. ☐ Put the possible input on the horizontal axis and the response on the vertical. ☐ Plot every pair. ☐ Examine the pattern and investigate clusters or unusual points.
Why it works
Relationships are difficult to judge from two separate tables. Plotting each pair as a point makes direction, strength, clusters, curves and unusual observations visible.
Look for…
Positive or negative association, curved relationships, separate clusters, changing spread and influential outliers. Check whether a third factor could be driving both variables.
Watch out…
Correlation is not causation. Do not pair data from different populations or time periods, and do not draw a straight line through a relationship that is visibly curved. A narrow data range can also hide a real association.
Worth noticing…
The lonely point away from the cloud may be an error, an exception or the most useful case in the set. It deserves checking before deletion.
Try next…
Use Regression Analysis when the relationship needs quantifying. Use a Fishbone Diagram to identify other plausible factors. Use Operational Definition when either measure is inconsistently recorded.