Dear Experts,
We're doing a study on a group of patients with the same disease, all of whom got the same treatment for three months. Some of them improved (thus becoming "responders"), others didn't ("non-responders"). They got MRI's at baseline (tp1: before the treatment) and again afterwards (tp2). All subjects received both MRI's and the interval between MRI's was three months for everyone.
Our goal is to see what changed in the brains of the responders and the non-responders from tp1 to tp2. Specifically, we want to compare the responders at tp2 to responders at tp1, non-responders at tp2 to non-responders at tp1 and we want to compare the change in responders to the change in non-responders. We'd like to create one GLM in which we can do all three of these analyses.
To do so, we're currently using the tutorial on Paired Analysis ( http://surfer.nmr.mgh.harvard.edu/fswiki/PairedAnalysis), with two groups (non-responders and responders), and two covariates (age and gender).
My first question is about the first FSGD file, containing the matching pairs listed consecutively. Should these subject names refer to the directories we got from the longitudinal processing stream (e.g. OAS2_001_MRI.long.OAS2_001) or those we got from running the 'normal', non-longitudinal recon-all command (e.g. OAS2_001_MRI)?
My second question is about the contrast file we should use to compare the subjects from tp2 to tp1 in each group separately. In the mailing list and in the tutorials, I'm mostly seeing examples to compare two groups with each other (with contrasts like '1 -1 0 0 0 0). But am I correct in thinking that to compare responders (the second group listed in the second FSGD file) at tp2 to tp1 while adjusting for gender and age, I should use the following contrast?:
0 1 0 0 0 0
Kind regards and thank you, M.C. van der Plas Leiden University Medical Center