Open ElanHR opened 9 years ago
Removal of mean connectivity could be a way of mean-centering the data. Removal of global signal is like removing the background signal, since all of the nodes share that signal. I would think that this increases the contrast overall.
The Shehzad et al paper showed they were robust against the removal of nuisance covariates (mean connectivity, global signal, motion).
Motion(of a person's head during a scan) being a nuisance covariate makes sense to me since it falls into experimental error/noise but the other 2 seem odd to characterize as nuisance variables since they seem to be inherent properties of the network. What does it mean exactly to remove them from a measurement? Am I looking at this the wrong way?