Dear FSL experts,
I've a doubt with the GLM and contrast that I should use after running concat ICA and dual regression. Imagine I have 4 subjects with three runs each. 2 subjects belong to group A and two subjects belong to group B (my sample is bigger, but I simplify here). I'm not interested in the difference between runs (its resting state), but just on the difference between groups A and group B.
I entered each run as a separate input in the group concat melodic ica analysis, thus I have 12 inputs. After it, I run dual-regression without randomise. And then is when my doubts appear, because there are at least two possible ways of managing my multiple runs data. I've tried both but i get different results, and I would like to know which is the optimal:
1. I create my general linear model (.mat) by specifying that the three first input are subject 1 (first column), the next three inputs are subject 2 (second column) and so on. Then in the contrast (.con) , if I want to compare the two groups, I just need to specify 11-1-1 (for comparison A-B) and -1-111 (for comparison B-A) (see the figure attached). If I do this, I obtain results that fit with which It should be expected.
2. Before running randomise, I create an average image of the three runs of each subjects (stage2 output from dual regression), Thus I will have only one image per subject. I also tried this, but the results are strange.
In both cases, I didn't modified the first column of the EVs in the GLM setup (the Group column), is that ok?
any help will be very welcome,
thanks in advance,
Diana
|