Confounding Variable
An unmeasured or unmodeled third variable that influences both the independent and dependent variables, creating a spurious association.
What is Confounding Variable?
A confounding variable is a specific type of extraneous variable that correlates with both the presumed cause and the presumed effect. If not controlled for, it distorts the estimated relationship, making it appear that an effect exists when it doesn't, or masking a true effect.
Why Confounding Variable Matters
Confounders are the primary threat to internal validity in observational research. They make it difficult or impossible to determine whether observed relationships are causal.
Example
A study finds a positive correlation between ice cream sales and shark attacks. Temperature is a confounding variable; hot weather increases both ice cream consumption and swimming in the ocean.
Common Mistakes
- Confusing a confounder (which causes both X and Y) with a mediator (which is the mechanism by which X causes Y).
- Assuming that statistical matching or regression can perfectly eliminate confounding, especially when there is unmeasured confounding.