Key takeaway: A confounding variable is a third factor related to both the explanatory and response variables, mixing their effects together so that the true cause of a change cannot be separated.
Grade 12 – AP StatisticsAP Statistics Unit 1Statistics and Data
Definition of Confounding Variable
A confounding variable is hidden machinery behind an apparent connection. It is linked to the explanatory variable and also influences the response, so a pattern between the two may come from the confounder instead of any direct effect.
Confounding is the main reason observational studies cannot prove causation. Experiments fight it with random assignment, which spreads the confounder evenly across groups, and with blocking or matching to compare like with like.
Formula and notation
Diagram the relationships with arrows:
\[ X arrow Z \to Y \]
where \(X\) is the explanatory variable, \(Y\) the response and \(Z\) the confounder. Association between \(X\) and \(Y\) can appear even when \(X\) has no direct effect on \(Y\).
Key facts
- A confounder must be related to both variables; a factor tied to only one cannot confound.
- Confounding differs from simple chance variation, which random sampling and large samples can reduce.
- Random assignment in an experiment breaks the link between the confounder and the treatment.
- Naming a plausible confounder is a standard way to criticize a causal claim.
Confounding Variable example
Fires attended by 6 trucks average \$480,000 in damage, while fires attended by 2 trucks average \$65,000. A columnist says sending trucks causes damage. Identify the confounding variable.
-
The explanatory variable is the number of trucks sent; the response is the dollar damage.
-
Large fires call for more trucks and also destroy more property.
-
Fire size is tied to both variables, so it is the confounder, and it can explain the whole pattern.
The size of the fire is the confounding variable; the data do not show that trucks cause damage.
Common mistake
Watch out
A student says the number of trucks cannot be confounded with damage because both numbers are large when fires are big.
Why it is wrong
That is exactly what confounding looks like: a third factor drives both variables. The student confused the confounder with the response. To check, ask whether a third variable could change both. If so, the apparent relationship is open to challenge.
Related terms and lessons
- Experiment vs. Observational Study
- Correlation vs. Causation
- Sampling Bias
- Scatter Plot
- Correlation Coefficient
Browse the glossary: A–Z list · AP Statistics vocabulary · Statistics and Data