the least-squares line for the above data is + .414, as usual slightly higher than the least-deviations line. This tendency for the least-deviations line to be slightly shallower is largely due to its lesser sensitivity to isolated or unusually high y-values. Where correlation is low, this tendency is not noticeable. The lesser sensitivity to freak values in the dependent variable may be an advantage; it may sometimes obviate the need for arbitrary exclusion of isolated observations.