A strong correlation can tell an incomplete story.

Move the noise slider to see a relationship become harder to detect. Add one outlier to see how much a single observation can change Pearson’s correlation.

What you’re looking at

These are synthetic demonstration data, not measurements of people or the real world. Thirty observations follow y = 0.7x + 10 plus a deterministic noise sequence. The optional outlier is (95, −25).

Pearson’s r measures linear association. It does not establish causation. Selection, confounders, measurement error and outliers can all change the story; look at the plot as well as the summary number.