I am trying to determine which of a number of variables differentiate between 3 groups. The groups were created based on cluster analysis, using a number of scales called the Emotional Availability (EA) Scales, that measure relationship quality between mothers and their children. I want to see if the 3 groups (high EA, low mother-EA/average child-EA and low child-EA/average mother-EA) differ on a number of other variables, such as sex of the child, age of the child, age of the mother, mother's education, quality of home environment, etc. When I ran ANOVAs for each, many were significant. I then realized that due to the intercorrelations between the variables, it would be better to run a MANOVA. When I did this, the overall analysis was significant and I found that the same variables were important, but when I checked which of my EA groups differed from each other, none of them did. Does this mean that the different variables don't work together in the same direction, and it would be better to just run separate ANOVAs despite the intercorrelations?

Thanks a lot!