Define multicollinearity and name a method to detect it.

• A. Multicollinearity occurs when predictors are uncorrelated; detect with chi-square test.
• B. Multicollinearity occurs when outcomes predict each other; detect with Durbin-Watson.
• C. Multicollinearity occurs when predictors are highly correlated; detect with variance inflation factor (VIF) or tolerance.
• D. Multicollinearity is the same as heteroscedasticity; detect with normality tests.

Answer: (C)

Multicollinearity occurs when predictors are highly correlated with one another, which makes it hard to estimate each predictor’s unique effect. Detecting it with variance inflation factor (VIF) or tolerance directly measures how much a predictor’s estimated variance is inflated by the presence of other predictors. A high VIF (often above 5 or 10) or a low tolerance (below about 0.1–0.2) signals problematic multicollinearity. This approach fits because it targets the core issue: shared information among predictors that inflates standard errors and muddles interpretation. Other options don’t fit: Durbin-Watson assesses autocorrelation of residuals, not multicollinearity; chi-square tests aren’t standard for detecting multicollinearity; and heteroscedasticity concerns unequal error variance rather than inter-predictor relationships.