Dear experts,
I am studying DTI parameters in two populations (controls and patients) and I want to perform an ANCOVA using randomise to test for 1) possible interaction between the diagnosis and education level, and 2) if no interaction emerges group differences. Unfortunately, the two diagnostic groups have significantly different education levels.
Do I have to demean such covariate in this case? If I don't demean how should I interpret the possible resulting group differences? As far as I understood, without demeaning, possible differences may depend upon differences in the covariate. Indeed, I would demean in order to attribute eventual differences in DTI parameters only to the diagnosis and not to education.
Any help would be really appreciated :)
Ciao,
C
|