I am conducting a 2-way (i.e., memory type x simulation type, 2x3) rANOVA within subjects. Both memory type and simulation type are categorical variables; the DV is continuous. Both the main effects of memory type and simulation type are significant, as is the interaction term. However, when I look at the simple effects, all pairwise comparisons are also significant, even after Holm correction for multiple comparisons. It seems one of the 3 simulation types has more similar memory types than the other 2, but I have no way of statistically confirming this result other than comparing t-scores/p-values/mean differences. What is the best way to determine what factors are driving the interaction?
I'm currently using a mixture of SPSS and R.