 |
 |
Hi, You can as you suggest add a new regressor: demean across all the patient symptom scale values and then pad with zeros for the controls. However note that this will just reduce the residuals in the model fitting (and potentially boost z-stats), but won't fundamentally make any difference to the group-difference contrast, as both subsets of the new EV (controls and patients subsets) have equal (zero) mean. The only was you could use this information to remove any contribution that the symptoms make to the group difference would be if you were justifiably allowed to consider the controls as having (e.g.) zero score, in which case you might have the controls as zero, and NOT demean the patient scores before combining with the control zeros, then demean the entire regressor. This would give a difference answer from above and _may_ be what you want? Cheers, Steve. On 13 Mar 2007, at 22:06, Dost Ongur wrote: > Hello all, > I have scanned two groups of subjects: patients and controls. For the > patient group, I have a symptom scale that I believe may be > relevant, but > this scale does not apply to controls. > I would like to include this symptom scale as a covariate to remove > any > contribution it makes to the group difference. Is it acceptable to > create > a third EV and use demeaned numbers for the patients and zeros for > controls? It seems that approach assumes that the controls have the > patient mean. > Any thoughts would be appreciated. > Thanks, > Dost ------------------------------------------------------------------------ --- Stephen M. Smith, Professor of Biomedical Engineering Associate Director, Oxford University FMRIB Centre FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK +44 (0) 1865 222726 (fax 222717) [log in to unmask] http://www.fmrib.ox.ac.uk/~steve ------------------------------------------------------------------------ ---