I have a 2(condition) x 3(group) mixed-design study.
I’m interested in the effect of a continuous time-constant covariate (i.e., a unique value for each subject, similar to age) on the interaction (Condition by group). The covariate is already demeaned across all subjects. In other terms, I want to test if different values of my covariate has an impact on the difference between the two conditions, separately for the three groups.
I am looking at effects on fMRI neuroimaging contrasts, so I am using the FSL package. Now the way the GLM model is set up, each subject is an explanatory variable. Therefore, a covariate for which there the same number of classes as there are subjects and which does not change between observation (as would be the case with age), is a linear combination of the subject explanatory variables.. and the model fails.
Do you have any suggestions for how to explore the effect of my covariate on this mixed-model study? Is the use of a time-invariant covariate correct with this GLM? If not, which kind of model should I use instead