492 Chapter 14 Repeated-Measures Designs
From the bottom row of that dialogue box you can specify what plots you would like to
see, what contrasts you would like to run, and any descriptive statistics you want printed
out. Then click on OK to run the analysis.
An abbreviated summary table appears below. I have omitted entries in the table related
to Greenhouse and Geisser and related corrections to condense the table. Notice that SPSS
presents separate tables for Within-Subjects factors and Between-Subjects factors, though I
would prefer to have them combined into one table with appropriate indentations.
Notice that there are multiple error terms in the table. The Group effect is tested by the
Error term in the Between-Subjects table. Then Cycle and Cycle 3 Group are tested by
Error(Cycle), Phase and Phase 3 Group are tested by Error(Phase), and Cycle 3 Phase and
Cycle 3 Phase 3 Group are tested by Error(Cycle 3 Phase).
From the summary table in Table 14.12, it is clear that nearly all the predictions were
supported. The only effect that was not significant was the main effect of Groups, but that
effect is not crucial because it represents an average across the shock and the No-shock
phases, and the experimenters had predicted little or no group differences in the Shock
phase. In this context, the Phase 3 Group interaction is of more interest, and it is clearly
significant.
The presence of an interpretable three-way interaction offers the opportunity to give
another example of the use of simple interaction effects. We would have predicted that all
groups would show high levels of suppression of the shock trials on all Cycles, because
anticipated shock is clearly disruptive. On No-shock trials, however, Groups A-Band L-A-B
should show less suppression (higher scores) than Group A-A, and this latter difference
should increase with Cycles. In other words, there should be a Groups 3 Cycles interac-
tion for the No-shock trials, but no such interaction for the shock trials. The simple effects
are shown in Table 14.13. (In these tables I have left in the corrections based on Green-
house-Geisser, Huynh-Feldt, and Lower-bound solutions to illustrate how they are pre-
sented by SPSS. Whether or not we choose to implement the corrections does not affect