Dear SPMers,
I'm interested in performing connectivity analysis in a resting-state dataset, and for that I would like to regress out the confounding effects from the white matter, CSF and the global mean. For that, I used Marsbar and the images generated by SPM5 Segment to extract, from the preprocessed scans, the mean time course of the signal in those regions. I'm using the Modulated Normalized images for the WM, GM and CSF as ROI definitions when extracting the time series (Raw data option in Marsbar).
Checking the time courses, I noticed that they are highly correlated -- typically, the correlation coefficient between any pair is > 0.9. In hindsight, that seems to be not so surprising, since the three time courses are averages taken across relatively broad areas of the brain. But I would like to make sure that this is the case, and that using them as covariates of no interest in my analysis is a sound procedure.
Also, because they are highly correlated, should they be made orthogonal before being used as covariates? And does the average time course of the gray matter serve as good proxy for the global mean?
Any advice will be greatly appreciated.
Thank you very much.
Eiji
|