4 x 2 ANOVA

A homework problem:
I must create a dataset for a 4 x 2 anova with at least 20 cases. The main effect for each way must NOT be significant, but the interaction effect MUST BE significant.

I have tried the following: 4 (school campuses) x 2 (grade levels) and score.
I have gotten datasets that have significance for one or the other of the ways, or no significance, but can't accomplish the interaction effect. I am using SPSS v.15 Grad Pack. 6 cases for each campus, 3 per grade level.

I am not sure how to adjust the scores so that the interaction effect is significant. Should I adjust the scores by campus--all scores from campus 1 are lower than the others; another adjustment could be to just adjust all the grade levels, i.e., all the fifth graders scored better...

Would anyone know an applicable principal to apply to this? Or have some suggestion as to what to manipulate? I am just using the trial-and-error method, but don't think this is the real way to learn what to do.

Oh, and there's a second part--dataset in which both ways are significant, but no significant interaction effect.