Dear SPM experts,
I intend to set up a seeded connectivity model. The first-level model will feature the seed-time-series as well as covariates of no interest to control for task, global, csf, wm and movement related variance. My question is how I should define the seed-time-series. In the past I've just extracted it from the first-level activity model (Task A > Task B, includes only task and movement as covariates) and only adjusted for the task related variance. However, this time series will still be confounded with global, csf, wm and movement related variance. I'm wondering if I'm not making a mistake when I use this time-series as my regressor of interest. It seems to me that when I include all the covariates I mentioned in the seeded connectivity model as described above I'm effectively correlating the seed-time series that is only adjusted for task related variance with the time-course for all voxels after adjusting for all covariates (i.e. by regressing out this share of the variance). That doesn't seem right to me. Hopefully somebody can help me out with this one.
Thanks for your help!
Felix Schmitz
|