Causation

First taught in High school statistics

Definition

Causation means a change in one variable directly causes a change in another variable.

Hot, sunny weatherIce cream sales ↑Sunburns ↑correlatedbut one does not cause the other
Ice cream sales and sunburns rise together, but hot weather causes both.
Advertisement

Causation is stronger than correlation. Two variables are correlated when they tend to change together. That alone does not show that one causes the other.

Correlation does not imply causation. A pattern can appear because:

  • a hidden (lurking) variable affects both, like hot weather affecting ice cream sales and sunburns;
  • the cause runs the other way;
  • the pattern is a coincidence.

The usual way to show causation is a controlled experiment. Subjects are put at random into a treatment group and a control group. If the groups then differ, the treatment is the likely cause. An observational study can show correlation, but on its own it usually cannot prove causation.

Advertisement