 |
 |
Hi Evan, You can run TBSS as usual, treating each timepoint as it it were a subject. At the far end, compute the subtractions and run a GLM to investigate the association with the differences in cytokine levels. It is also possible to, instead of subtractions, assemble the model of the GLM as a paired t-test, including (and testing) an extra EV that is the cytokine level at each of the two MRI acquisitions. Remember to use exchangeability blocks, defining one EB per subject. All the best, Anderson On 1 January 2016 at 21:54, Evan Stone <[log in to unmask]> wrote: > Hey all, > > I'm a but rusty (I haven't touched FSL/TBSS in over a year) but finally a > project I'm working on has enough data to analyze. We have 24 subjects with > baseline DTI and +1 year DTI, for a total of 48 separate scans. I would > like to be able to use individual differences over the course of the year > to correlate with blood levels of specific cytokines that we drew at the > time of the MRI acquisition. That being said, what would be the best way to > analyze this in TBSS? > > Currently my plan is to pull everything through the TBSS pipeline and > split the contrast groups back into individual, spatially-normalized > skeleton FAs, then subtract an individual's skeletonized 1y from baseline. > That way, I can pull FA values on whole-brain and tracts from atlases; > however, statistical significance in differences between an individual's > baseline and 1 year group seems to be challenging from this route. > > I know this is a question not only about TBSS but about statistics, but > any direction would be greatly appreciated. > > Thanks! > -Evan >