Skip to content

Forcing a return line through the origin

Deleting the intercept changes the least-squares problem and can invalidate familiar fit identities.

On this page

State which line is fitted

The public tool fits Y=a+bX. A through-origin model forces a=0 and estimates slope from uncentered sums. That restriction should come from the modeling task, not from an attempt to remove an inconvenient fitted intercept.

Inspect the original mean context

In the original record, x̄=0 and a=1.2%. A through-origin line cannot pass through the centroid (0,1.2%). Its residual sum is not forced to zero by an intercept normal equation.

Keep the statistic definition attached

The standard intercept-fit R² and its equality to squared correlation should not be exported unqualified to a no-intercept calculation. A different fitting criterion or score definition answers a different question.

Retain a supported restriction

If zero intercept is required, document that model and its appropriate formulas separately. The tool does not silently switch between those models or suppress the intercept.

Further references