Causal Confounding

“Two groups may differ before the supposed treatment in a way that also changes the outcome.”

01 / PLAINLY

What it means, plainly

Confounding occurs when a prior difference influences both the exposure and the outcome. Part of the comparison credited to the exposure may then come from that shared cause.

02 / CONTEXT

A little more

Confounding occurs when a common cause, or equivalent structure, mixes the exposure's effect with baseline differences. Adjustment can help only when the needed variables are properly identified and measured.

03 / WHY IT MATTERS

Why it matters

Recognizing it keeps an association from being read as the effect of one thing on another without checking how the groups formed.

04 / EXAMPLE

A familiar example

People taking a supplement also exercise more; comparing health without handling exercise mixes the two processes.

05 / LIMIT

What it does not mean

Adding every available variable to a regression can create bias when one is a mediator or collider.

06 / NOTICE

Notice it in your day

For the supplement example, draw arrows from exercise to supplement use and health. Explain which comparison becomes mixed.

FINAL NOTE

The idea worth keeping

Choose adjustments from a causal model, not an automatic covariate list.

QUESTIONS / 02

Questions people still have

What does a confounding cause do?

It influences both exposure and outcome, mixing baseline differences with the effect being estimated.

Should every available variable be adjusted for?

No. Some may be mediators or colliders and can introduce bias.