Residual
Also known as: OLS residual, Fitting residual
A residual in OLS is $e_i=y_i-\hat\alpha-\hat\beta x_i$ — the gap between observed $y_i$ and the fitted value. Residuals estimate the unobservable errors $\varepsilon_i$.
RSS $=\sum e_i^2$ is the least-squares criterion value. Residual diagnostics (plots, Durbin–Watson) check homoscedasticity, dependence, and nonlinearity. Residual df is typically $n-p$.