 |
 |
Dear Darren, >Dear SPM: > >I want to set up an fmri design which will basically just allow me to >remove a variety of covariates of no interest. Once these covariates >are removed from the data I will use the data in another way. Karl >answered a previous question about how to remove the covariates using >spm_regions. This question is one step before that. > >What I want to know is how do I set up a design matrix so there are >basically no covariates of interest? spm97 allowed one to do this by >answering 0 to the covariates of interest question. I don't know how >to do this in spm99. > >The following did not seem to work > >I set up a design as follows. > >There are 200 scans in the session taken at a TR of 3. > >So # of trials or conditions= 1 >SOA of trial 1 = 200 >Onset of trial 1 = 0 >Length of trial 1 = 200. >not stochastic, fixed onset. > >I then enter a number of covariates which end up being my covariates >of no interest. > >The resulting design matrix has all 0's as it's first column, >followed by the covariates [of no interest] followed by a mean vector >of all 1's. > >I then looked at a series of F contrasts- >The first column of the design (all 0's) is not uniquely estimable so >I can't define a contrast as just 1. Likewise choosing columns for >the reduced design of either 1 or 2:112 doesn't give the right answer. > >Any ideas? > Why don't you try and insert a couple of events/epochs in your covariate of interest (they need not have any relation to any actual events/epochs). That should ensure estimability, and allow you to use the 1 0 ... 0 contrast. From the prespective of obtaining adjusted data, it shouldn't matter at all what you have in your covariate of interest (as long as it is estimable). >darren >Darren R. Gitelman, M.D. Good luck Jesper Jesper Andersson Wellcome Dept. of Cognitive Neurology 12 Queen Square London WC1N 3BG phone: 44 171 833 7484 fax: 44 171 813 1420 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%