In a MANOVA, is it problematic if the correlations between the DVs are different across cells (across each combination of the IVs)? In other words, is it ok if the correlation between Y1 and Y2 is changes depending on the levels of X1 and X2. I am asking this because there seems to be a contradiction between the following two points:
1) On the one hand, MANOVA assumes homogeneity of covariances between DVs across all groups. So if my interpretation of this assumption is correct, given that covariances and correlations are closely related, the assumption means that the correlations between DVs should be roughly the same across all groups (cells).
2) On the other hand, I read that MANOVA can be useful to examine the effect of IVs on the relationships between DVs. If my understanding of this is correct, then variations in the correlations between DVs across groups is expected and an object of study for MANOVA.
In my current understanding, points 1 and 2 are contradictory. Clarifications on these issues and on the main question would be much appreciated!