Dear all,

I am running a RM ANOVA with 2 independent variables (factor A with 3 levels, factor B with 5 levels). The AxB interaction is significant and I want to know using a post hoc test in which level of the factor B the factor A had a significant effect on the dependent variable.

In SPSS' Syntax, I write "/EMMEANS=TABLES(A*B) COMPARE(A)".

This leads to :

• Pairwise Comparisons, which indicates comparisons of each pair of levels of the factor A (1v2, 2v3, 1v3) in each 5 levels of the factor B. In this case, I can apply a Bonferroni correction by subdividing the significance threshold by 15 (3 levels of factor A * 5 levels of factor B).
• Multivariate tests, which indicates in each level of the factor B if the factor A had a global effect on the dependent variable (this gives me 5 results, one for each level of the factor B).

My question is about this second point: Are these tests strictly speaking post hoc tests? To me, they look more like ANOVAs performed on each fixed level of the factor B.

If they are indeed post hoc tests, can they be corrected using Bonferroni's correction? If so, do I have to divide the significance threshold by 5 since I make 5 comparisons in total?

I joined the two related SPSS outputs. In the multivariate tests output, the simple effect of A is significant in each level of B. If I use a Bonferroni correction, it will not be significant at B's level 1. What is the correct thing to do?

Thanks,

Gauthier