Ordinary least squares
A residual is the observed response minus its fitted value. Squaring avoids cancellation of positive and negative residuals and gives large errors greater influence. Least-squares estimates do not automatically make model assumptions true; inspect residuals, outliers and the business context, and validate on suitable new data.
Worked example
For residuals of −2, 1 and 3 units, the sum of squared residuals is 4 + 1 + 9 = 14 squared units. OLS chooses coefficients that minimise this quantity for the specified model and data.
Common mistake
Minimising the sum of signed residuals instead of squared residuals, or interpreting a fitted association as causal.
Where it appears in your exams
| Qualification | Exam | Relationship | Period |
|---|---|---|---|
| ACCA | MA | Supporting | September 2026–June 2027 |
| FIA | FMA | Supporting | September 2026–June 2027 |