Confounding Variable

mechanism

A hidden third factor can produce both variables, creating a correlation without a direct causal link. Causal analysis must test plausible common causes before drawing an arrow from one observation to the other.

People who sleep with their shoes on may wake with headaches. The shoes did not cause the pain: being drunk made both outcomes more likely.

The missing fork in the causal path

A confounding variable sits upstream of two observations. It changes each through a separate path, so the observations move together even though neither produces the other. If you examine only the pair, the shared movement tempts you to draw a direct arrow between them; the real structure is a fork, with the hidden factor at its base. That is why correlation is evidence to investigate a mechanism, not proof that you have found one.

Where it shows up

Shoes, headaches, and alcohol

Sleeping in shoes predicts a morning headache only because intoxication can lead to both. Once the common cause is identified, the apparent shoes-to-headache mechanism disappears.

Not every wrong arrow hides a third variable

Confounding is only one rival explanation for a correlation. The arrow might run in reverse, operate in both directions, or reflect coincidence. Naming a plausible third factor does not establish it either; you still need evidence that it influences both observations.

Draw the fork before the arrow

When someone claims that X causes Y from a correlation, write X and Y on paper and add a blank node above them. Ask what concrete condition could independently produce both—and what comparison would hold that condition constant. If the direct link vanishes under that comparison, the blank node was doing the causal work.

Episodes that teach this

The Great LIE - Correlation is Not Causation - FutureIQ · explained at 8:09 7,309 views
Sleeping with shoes and morning headaches are both caused by being drunk.