Dear users,
here it is the issue:
My aim: within a cross-sectional study, I want to compare at voxel level the age-related trajectory of GM volume between two diagnostic groups. (I consider as GM volume at each voxel belonging to the segmented GM map, its modulated intensity.)
My method: to perform an ANCOVA through a full factorial design, where "Diagnosis" is the fixed factor and age the covariate. Within this design I can easily check for the eventual interaction between fixed factor and covariate and for the presence of a main effect.
My problem: I know that the first hypothesis of ANCOVA is the linear dependence between the dependent variable (GM in my case) and the independent one (age). Within the full factorial described above I assume the presence of such linear relationship. However, if I check it performing a multiple regression separately for each diagnostic group, I do not obtain significant results.
How should I proceed now that I know the basic hypothesis of my design is not fulfilled? Would it be correct to check for the presence of a linear relationship between variables a posteriori, only in the clusters resulting, for instance, from the main effect?
Any kind help would be appreciated :)
Chiara
|