Dear Experts,
I'm attempting to basically follow the outline from your paper on "Group comparison of resting-state FMRI data using multi-subject ICA and dual regression" in order to analyze data from an event-related paradigm.
My questions mainly concern Post-Stats setup in Melodic.
I'm planning to use multi-session temporal concatenation, rather than tensor. I have timing info for some, but unfortunately not all, participants, and I think timing inexactitude is one of the main reasons that I haven't been able to do this analysis in FEAT.
Question 1) Is it then worthwhile to enter a design.mat file in the "Timeseries model" section? Would I use some kind of average/approximate timing creation? Also, I had been using 3-column format files in FEAT - would I do the same setup for use in Melodic? My understanding is that this would possibly be useful in determining which Components are related to my design.
Question 1a) I've attempted to use a design.mat file in a Single-session ICA run, just to see if it made a difference, but there seemed to be no difference in the output - I didn't get a plot of the total model fit, GLM table describing the fit, F-tests, etc.
Question 2) For "Timeseries contrasts": If I'm planning to do group-level contrasts, and I only have 1 EV on an individual basis, should I enter anything here?
Question 3) For "Session/subjects model & contrasts": I'm looking to do group-level contrasts between two groups of about 20 participants each. Do I enter anything into this section in Melodic, or later on when I run Randomise on the concatenated 4D data? Seems like I have opportunity to do it in both places, but I don't want to be redundant.
Question 4) This regards the details of how I'll actually run this analysis: Will I be running Melodic twice, once for each group? Or am I able to run it just once and use the Post-Stats section to tell Melodic that there are two separate groups (and which participants belong to each group)? Which method would make it easiest to complete the process, ie later using Randomise?
I may be mistaken on my understanding of some of this - I'm new to the whole process. Thanks in advance for any response.
Charlie
|