standard error of the estimate
a measure of the actual variability about the regression plane of the 
underlying population.  The underlying population generally falls 
within about two standard errors of the observed sample.