The residual values (or residuals) of a linear model are equal to the actual values minus the predicted values. For non-linear models, the definition is more complicated, but in all cases smaller residual values correspond to better model fit. Residual analysis can help evaluate the overall quality of a model and identify outliers.
To create a numeric column out of a model’s residual values:
Because the values are copied, changes to the underlying model will not affect the values in the new column.