We have seen that the one-at-a-time approach to investigating the effect of
two factors on a response is suitable only for situations in which the two factors do
not interact. Although this was illustrated for the simple case in which two factors
were to be investigated at each of two levels, the inadequacies of a one-at-a-time
approach are even more salient when trying to investigate the effects of more than
two factors on a response.
Factorial treatment structures are useful for examining the effects of two or
more factors on a response y, whether or not interaction exists. As before, the
choice of the number of levels of each variable and the actual settings of these variables
is important. However, assuming that we have made these selections with
help from an investigator knowledgeable in the area being examined, we must decide
at what factor–level combinations we will observe y.
Classically, factorial treatment structures have not been referred to as designs
because they deal with the choice of levels and the selection of factor–level
combinations (treatments) rather than with how the treatments are assigned to
experimental units. Unless otherwise specified, we will assume that treatments are
assigned to experimental units at random. The factorial–level combinations will
then correspond to the “treatments” of a completely randomized design.