On this page
Retain the original sample identity
The five-period record ends with benchmark 2% and asset 6%. The full fit has slope 2.1, intercept 1.2% and residual sum of squares 1.2 percentage points squared.
Construct a declared four-period comparison
Leaving out that final pair gives slope 1.8 and intercept 0.9%, with residual sum of squares 0.3. The sample has changed, so the smaller residual sum is not automatically a fair model-quality improvement.
Inspect why a point can matter
The final benchmark value lies away from the full-sample mean and contributes to the explanatory variation. Its influence should motivate review of period identity, source accuracy and model context, rather than an unsupported outlier label.
Keep legitimate observations
A valid unusual observation can contain important information. Document a correction or exclusion only when supported by the actual data issue or estimation design. The calculator does not automatically delete observations.