Residual sum of squares

Also known as: RSS, Sum of squared residuals, SSE

Residual sum of squares (RSS) is $\sum_{i=1}^n e_i^2$, the sum of squared OLS residuals. Least squares minimises RSS. The unbiased error variance estimator in simple regression is $s^2=\mathrm{RSS}/(n-2)$.

$R^2$ $=1-\mathrm{RSS}/\mathrm{TSS}$ compares RSS to total sum of squares. Lower RSS means better in-sample fit — not automatically a better causal model.

← Back to CS Glossary