Data = Fit + Residuals, or
Data = Smooth + Rough
As Hartwig explains:
"The difference between an observed value and a smoothed value is
known as the residual value, the importance of which is shown by the
following rearrangement of terms in the basic equation for the ex-
ploratory approach: rough = data - smooth. In other words, a set of
residual values is the same as the rough, and because exploratory data
analysis is an iterative process in which successive roughs are examined
for evidence of additional smooth, the examination of residuals is a
crucial part of the approach."