Dear Liron

Regarding the 'effect of rest' covariate, if you plot its timecourse you will see it basically contains a small ramp at the beginning of the scan. For resting-state analyses it is perfectly fine not to include this effect as part of your modeled potential confounds (but, on the other hand, some times doing so may protect you against some potential ramping effects in your data -e.g. if you have not already removed the first few scans from your acquisition-). See this thread (http://www.nitrc.org/forum/message.php?m...) for a bit more information. 

Regarding your second question, I am not sure I understand your question correctly. Just to clarify, the aCompCor procedure, as defined in the original Behzadi paper, involves regressing out noise components estimated from the White matter and CSF regions, which is done simultaneously in a single regression model also removing any other standard nuisance parameters that one would typically use to characterize the BOLD signal (e.g. task, realignment, detrending, etc.). If you leave the 'White matter' and 'CSF' effects in the 'confounds' list, that will regress-out those components from the BOLD signal (any effect that you include int he 'confounds' list will be regressed-out from the BOLD signal). If you are not including the movement parameters there because your data has already been denoised using different software, that is typically perfectly fine (but there may be some issues with this approach such as the regBP vs. BPreg issues discussed in Hallquist's "the nuisance of nuisance regression" paper)

Hope this helps
Alfonso

Originally posted by L R: