Dear Christy,
There are certainly folks more expert than I on this issue, but since I
haven't seen an answer, I'm giving my limited 2 cents (I expected to be
corrected where I'm wrong):
At 06:39 AM 8/1/2005, Christy Marshuetz wrote:
>% design (user specified covariates)
>%---------------------------------------------------------------------------
>for ses=1:nses
> SPM.Sess(ses).C.C = [1:208]'; % [n x c double] covariates
> SPM.Sess(ses).C.name = {'Regressor (ses)'}; % [1 x c cell] names
>end
>
>Several questions:
>1) Is the regressor indeed modeling linear drift? If not, what?
Mostly people generally specify movement parameters here (if anything), but
maybe you're modeling within-session drift. spm_detrend (as called during
glm design) removes linear drift without needing to explicitly specify a
drift regressor, but I believe that this is drift across sessions.
>2) Is it redundant (colinear?) with other variables already in most SPM
>models (e.g., intrinsic autocorrelations?)?
Hard to say since we're not sure what it is modeling.
>3) Can or should I try eliminating this extra regressor?
You can.
>4) If yes to 2, how do I go about doing so? Simply deleting it, or
>putting in [none] doesn't work.
To eliminate it, replace your text with the following:
for ses=1:nses
SPM.Sess(ses).C.C = []'; % [n x c double] covariates
SPM.Sess(ses).C.name = {}; % [1 x c cell] names
end
Hope this helps.
Cheers,
Charlotte
___________________________
Charlotte A. Boettiger
University of California, San Francisco
Ernest Gallo Clinic & Research Center
5858 Horton Street, Suite 200
Emeryville, CA 94608
(510) 985-3973 - phone
(415) 336-2055 - mobile
|