Chapter 7, Table 15: Nonorthogonal Factorial Designs via SPSS Syntax

The following hypothetical salary data represents a nonorthogonal two-by-two factorial design. The first factor (sex) is crossed with college (degree or no degree). The primary question of interest is whether or not there is sex discrimination in terms of salary.

As before with factorial ANOVAs, this analysis makes use of the UNIANOVA procedure. What is important to notice here is the /METHOD=SSTYPE(3) option is included. Whereas before it was not necessary to include such a specification (because cells always had an equal number of participates and thus all types of the sums of squares were equal), for nonorthogonal designs it is import to realize the type of sums of squares being used (the differences are formally given in Tables 7.19 through 7.22).

Notice that the type of sum of squares is selected by specifying 1, 2, 3, or 4 in the SSTYPE() option (Type 3 sums of squares is the default in SPSS). Note that for the effect entered first into the Fixed Factor(s) box can make a difference. For example, if the effect “A” is entered into the first effect entered into the Fixed Factor(s) box, the analysis actually performed is a “Type I sums of squares where ‘A’ is entered into the equation first.”