 |
 |
Dear Joe, "Jose' Ma. Maisog" wrote: > When one performs a multisession fMRI analysis with SPM, in either its > SPM96 or SPM99 incarnation, my guess is that SPM removes intersession > effects, perhaps behind the scenes. If this is so, if each session were > a different subject, and if one tried to use the subjects' age (rounded > to the nearest year) as a confound in the analysis, would this result in > a problem of linear dependence in the columns of the design matrix? > Yes and no. It would result in a linear dependence in the design matrix (there will be a linear combination of "session-regressors" that equals your "age-regressor"). It will not be a problem as long as all you want to do is to use it as a confound (since SPM uses the generalised inverse), but then again it would not do you any good either since the space of the confounds is identical in both cases. You cannot use it as a covariate of interest since it is exactly in your confound space. In short, the confounds already modelled removes any age effects, and more. > > Thanks, > > Joe. Good luck Jesper %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%