Chapter 7, Table 11: Interaction Contrasts for
Factorial Designs via SPSS Syntax
The following data represents a two way factorial design
where diagnosis and task are fully crossed, each with three levels. Of interest
for the results displayed in Table 7.12 is whether the interaction contrast
specified in Figure 7.3 and 7.4 is statistically significant. Namely the
question pertains to whether the relationship of the mean of grammar and classification versus recognition
differs for those in the amnesic and
The /CONTRAST statements accompanied by a SPECIAL matrix is what defines the contrasts that are to be carried out. Notice the first row is (generally) always filled with as many 1s as levels of the factor specified in the parentheses following the /CONTRAST statement. Excluding the first row of 1s, there are the numbers of levels of the factor minus one remaining rows. These rows should (generally) always specify orthogonal contrasts, otherwise the results will differ from what is expected.
Because interest lies in an interaction contrast, we must specify a SPECIAL matrix for both factors (the first row in each SPECIAL matrix is what is of interest for the present example). The /DESIGN command specifies the particular contrast that is desired. Because we are interested in an interaction contrast, we will be crossing the individual coefficients from the first row of each of the SPECIAL matrices. We do this by calling upon the first row from the diagnose SPECIAL matrix and crossing it, using BY, with the first row of the task SPECIAL matrix.