The error term for a single data point is written as the observed y minus the:

1. Residuals are measured vertically from the point to the line. 2. Each data point has the form (x, y) and the corresponding point on the line has the form (x, y-hat). 3. The term y0 minus y-hat0 is called the error or residual. 4. So the subtraction is the observed value minus the predicted value at the same x. 5. Subtracting the mean instead would give a deviation, which is a different quantity. _Source: OpenStax Introductory Statistics 2e (CC BY 4.0), Ch 12 'Linear Regression and Correlation', sections 12.1-12.6_