Dear FSL experts,
I have some basic questions about setting up the analysis for my melodic ICA analysis. I have 3 groups, controls, patients and relatives with 35 subjects in the first two and 20 relatives.
I first ran an ICA on each group separately while not specifying the number of components. I ended up with 18 components with control and 12 for the patients and relatives. First, I would like to ask you what is the significance of the difference in the number of components? Second can I use this kind of data to compare between subjects or do I have to put all subjects together when I run the ICA?
Alternatively I am planning to combine subjects from two groups (say patients vs control) and a do a group ICA. My design matrix will specify the two groups and I will specify contrasts. Then of course I will do a dual regression. Would it be better to put the three groups together and then specify the design matrix and contrasts? In that case how should these look like? Finally, do you recommend setting up automatic ICA components or to specify a certain number?
Best regards and thanks in advance,
George
|