In statistical models, confounding is said to occur when the apparent dependence of the response on a predictor is partially or wholly due to the dependence of both on a third variable not included in the model, or dependence on a linear combination of other variables included in the model. Confounding with a variable included in a model is often called multicollinearity. A synonym is *aliasing*, used in design of experiments.
In statistical models, confounding is said to occur when the apparent dependence of the response on a predictor is partially or wholly due to the dependence of both on a third variable not included in the model. The causal relation, if any, between predictor and response is thus obscured. Observational studies are especially prone to confounding, the only remedy being to include all potential confounders in the model. Experiments mitigate confounding through randomization; though when randomization is restricted by blocking, some main effects or interactions may be partially or wholly confounded with blocks.
