Multicollinearity

Multicollinearity is high correlation among the predictor variables in a regression model. The model as a whole may still predict well, but the individual coefficients become unstable and difficult to interpret, because overlapping predictors are competing to explain the same variance.

The classic symptoms are a significant overall F-test with no significant individual predictors, coefficients that flip sign or change sharply when a variable is added or removed, and implausibly large standard errors.

Diagnose it with the Variance Inflation Factor — above 5 signals concern and above 10 is generally treated as serious — or its reciprocal, tolerance below 0.10. Remedies include dropping one of the redundant predictors, combining them into a composite or factor score, centring variables before building interaction terms, or moving to ridge regression.