Dear SPM/VBM users!
I am using the VBM8 toolbox in SPM8 to analyse difference in grey matter volume between two groups of patients, using four covariates of no interest. I want to add a fifth variable that is strongly associated with, and not independent from the factor and analyse the effect of this variable as well as the effect of the factor.
For example: Comparing drunk persons with sober persons, adding "being careless" as a binary covariate (drunk persons are more careless and careless persons drink more)
If I should add the variable to a two-sample T-test it would take away most of the variance caused by being drunk. I think I need to use the full-factorial to set up an ANOVA with two factors with two levels but is there any way to model the dependence between factors? Should I choose "Independence - No" for both factors or just for factor 2? Do something else?
I believe I should assume unequal variance for both factors. Should I?
The covariates of no interest are independent of the factors. Even if I choose "no centering" for the covariates, in the Anova analysis SPM still seems to center covariate values around group (factor+level) means, at least for groups with few subjects (the group is grey in the design matrix). Is it so, or why is it grey?
I would be most grateful for your help!
Best Regards
Carl Johan Ekman
|