Correlation, Causation, and Confounders

One of the most important reasoning skills tested in the UCAT Decision Making subtest is the ability to distinguish between correlation (two things occurring together) and causation (one thing causing another). Many UCAT questions present data showing a statistical association and ask whether a causal claim is justified. This lesson teaches you to identify confounders, evaluate causal claims, and recognise common logical errors in data-based arguments.

---

Correlation vs Causation

Correlation

Two variables are correlated when they tend to change together. This can be:

Type	Description	Example
Positive correlation	As one increases, the other increases	Ice cream sales and sunburn rates
Negative correlation	As one increases, the other decreases	Exercise frequency and body fat percentage
No correlation	No consistent relationship	Shoe size and exam results

Causation

Causation means one variable directly brings about a change in another.

"Smoking causes lung cancer" — there is a direct biological mechanism by which smoking damages lung tissue and leads to cancer.

The Critical Distinction

Correlation does not imply causation.

Two variables can be correlated for several reasons:

1. A causes B (direct causation)
2. B causes A (reverse causation)
3. C causes both A and B (confounding variable)
4. Coincidence (no real relationship)

---

Confounding Variables

A confounder (or confounding variable) is a third variable that is associated with both the exposure and the outcome, creating a spurious apparent relationship between them.

Example

Observation: People who drink red wine moderately have lower rates of heart disease than non-drinkers.

Tempting conclusion: Red wine protects against heart disease.

Confounders: Moderate wine drinkers tend to have higher incomes, better diets, more leisure time for exercise, and less overall stress. These factors — not the wine itself — may explain the lower heart disease rates.

How to Identify Confounders

Ask: "Is there a third variable that could explain both the exposure and the outcome?"

Observed association	Possible confounder
Firefighters at fires AND damage done	Size of the fire (more firefighters sent to larger fires; larger fires cause more damage)
Hospital size AND mortality rate	Case complexity (larger hospitals treat more complex cases)
Vitamin use AND longevity	Overall health-conscious lifestyle
Coffee consumption AND heart disease	Smoking (coffee drinkers in some populations are more likely to smoke)

---

Evaluating Causal Claims

When a UCAT question presents a causal claim based on data, use the Bradford Hill criteria (simplified for the UCAT) to evaluate it:

Criterion	Question to ask
Strength	How strong is the association? (A larger effect size is more suggestive of causation)
Consistency	Has the finding been replicated in other studies?
Temporality	Does the cause precede the effect? (Essential for causation)
Dose-response	Does more of the cause lead to more of the effect?
Plausibility	Is there a plausible biological/logical mechanism?
Alternative explanations	Have confounders been controlled for?

UCAT Tip: You do not need to memorise these criteria by name. But you do need to recognise that a single observational study showing an association is NOT sufficient evidence for causation.

---

Reverse Causation

Sometimes the presumed direction of causation is wrong.

Claim: Depression causes unemployment.

Alternative: Unemployment causes depression.

Or both: Unemployment and depression may reinforce each other in a cycle.