Dear Vladimir thanks again for your support. I'm sorry to disturb even more. I have some doubts on how to implement the procedure you suggested me. I can fit a model separately for each subject, therefore I will obtain 60 models one for each subject. What I should do next?
I have one clinical (22 subjects) and one control group (38 subjects). Let's image that I find the best 38 models, one for each control subject. My purpose would be to show that this model does not explain the data in the clinical group. For example let's image that in control group the best model across subjects finds a biderection relationship between superior parietal lobule and anterior cingulate cortex. I would like to test if in the clinical gruoup I have a different connection pattern. Based on you suggestion of running one model for each subject, how could I provide a response to my experimental question?
Secondly, if I fit a model separately for each subject, there is the possibility that the model that better explain the ROI interaction in one subject, may be different in another one. So when I contrast the models, there could be variability within the group, not only between groups. This is natural in any experiment of course, but I was wandering, is this a problem in the DCM approach? For example if the models are different across the subject in the control group how could I define the best model across subjects in the group?
I hope I expressed myself correctly
Thanks for your help
|