White Correction

Keeps the OLS estimates and replaces each unknown Var(ui)Var(u_i) with the squared residual u^i2\hat u_i^2 to get robust standard errors.

When to use

Use it when Heteroskedasticity is present but its form is unknown. Coefficients and R2R^2 do not change, but standard errors and tests do, and unlike Weighted Least Squares the estimator is not efficient.

See Lecture 12.

Appears in