The residuals are the difference between actual data values and their modelled values after some sort of data fitting process. For example in linear regression the distance of the data points from the regression line. In general, for data of the form (xi,y(sub>i) if a model 'M' has been fitted to predict y form x, then the residuals ri are given by:
ri = y(sub>i – M(xi)
Often the model fitting process is expressed in reducing the {[sum of squares}} of the residuals or some similar metric as a fitness function.
Used on pages 130, 134, 139