Which issue arises when features are highly correlated, making it difficult to disentangle their individual effects and causing unstable parameter estimates?

Which issue arises when features are highly correlated, making it difficult to disentangle their individual effects and causing unstable parameter estimates?

• A. Ridge
• B. Heteroskedasticity
• C. Outliers
• D. Multicollinearity

Answer: (D)

When predictors are highly correlated, you’re dealing with multicollinearity, which makes it hard to disentangle their individual effects and leads to unstable parameter estimates in a regression model. Because the predictors share so much information, the model can’t reliably attribute an observed change in the outcome to one predictor over another. This shows up as coefficient estimates that swing wildly with small data changes, large variances, and inflated standard errors, which also makes interpretation of each predictor’s unique impact unreliable. In practice you might see very similar predictors with coefficients that flip signs or become highly sensitive when you tweak the data. Ridge regression can help mitigate this by adding a penalty that shrinks coefficients and reduces variance, improving stability for prediction. But the root issue when features are so closely related is multicollinearity itself. Heteroskedasticity involves changing residual variance and affects efficiency, outliers are unusual points that can distort estimates, and those address different problems not tied to correlations among predictors.

When predictors are highly correlated, you’re dealing with multicollinearity, which makes it hard to disentangle their individual effects and leads to unstable parameter estimates in a regression model. Because the predictors share so much information, the model can’t reliably attribute an observed change in the outcome to one predictor over another. This shows up as coefficient estimates that swing wildly with small data changes, large variances, and inflated standard errors, which also makes interpretation of each predictor’s unique impact unreliable. In practice you might see very similar predictors with coefficients that flip signs or become highly sensitive when you tweak the data.

Ridge regression can help mitigate this by adding a penalty that shrinks coefficients and reduces variance, improving stability for prediction. But the root issue when features are so closely related is multicollinearity itself. Heteroskedasticity involves changing residual variance and affects efficiency, outliers are unusual points that can distort estimates, and those address different problems not tied to correlations among predictors.