When using time series data, it is always advisable to
plot the residual error terms against time. This is true even when time has not been
explicitly incorporated in the model.
If there is a correlation between the error term and time, it will become
immediately visible in this plot.
Figure (e) demonstrates a case where negative residuals
are associated with early observations and positive residuals with latter
observations. In such a case it may be
worthwhile to include time as a variable in the model.
