Ordinary least squares

Also known as: OLS, Least squares, Linear least squares

Ordinary least squares (OLS) estimates linear model parameters by minimising residual sum of squares:

\[\min_{\alpha,\beta}\sum_{i=1}^n(y_i-\alpha-\beta x_i)^2.\]

Closed form: $\hat\beta=S_{xy}/S_{xx}$, $\hat\alpha=\bar y-\hat\beta\bar x$. Under Gauss–Markov assumptions, OLS is BLUE. With Normal errors, OLS $=$ MLE and exact $t$ inference applies.

← Back to CS Glossary