Hi Anderson,
Thank you for the quick reply! I'm interested in looking at the unique connectivity of a seed region in the thalamus while controlling for the connectivity of two nearby thalamic nuclei. Because the regions in question are very small (<10 fMRI voxels each), there's a lot of concern about partial volume contamination, so this analysis is intended to test whether signals from these nearby regions are influencing the connectivity observed for our target region. We're hypothesizing that the connectivity pattern for the seed region in this multivariate analysis will be similar to the connectivity pattern from the bivariate analysis (indicating that the adjacent regions are not influencing the connectivity of our target region).
This analysis was originally conducted using the SPM Conn toolbox, so the inputs I have are the subject-level connectivity maps (in the form of 3D nifti files) for each of the three seed regions. Each of these contains the z-transformed correlation coefficients between the mean timeseries of the seed region and every voxel in the brain. So, our goal is to test the connectivity of the primary seed using TBSS while specifying the voxelwise connectivity values for these secondary ROIs as nuisance regressors.
The subjects are also evenly split into two groups, and a separate analysis is looking at the comparison between groups. However, for this analysis we're just interested in the whole-brain connectivity pattern across the entire sample, so we'd like to input this grouping variable as an additional nuisance covariate. I currently account for this in my model by setting EV1 as the target seed connectivity (+1 for all), EV2 as group (-1 or +1), and the contrast for positive connectivity as (1,0).
Does that clarify things? Please let me know if any other information would be helpful.
Many thanks,
-Ely
|