Dear Dave,
> I'm performing a second-level analysis using the PET/SPECT models
> interface in SPM99. I have 14 subjects, each with four conditions which
> have corresponding contrast images. I would like to ask about the
> difference between using the single subject - covariates and conditions
> option and one of the multi-subject options, as I am leaning towards
> using the Is the only difference that the multi-subject design will
> model subject x condition interactions explicitly (something that I am
> not too bothered about, as I am only interested in the second-level
> mean effects of each of my conditions)? I understand that this may give
> me a less sensitive model if subject x condition interactions are
> substantial.
In second-level analyses there should, generally, be only one contrast
per subject and these enter into a one-sample t test. If you enter
more than one contrast per subject into the second-level model then you
have to assume sphericity. You would normally only do this if you had
to (e.g. to perform conjunctions at the second level). Remeber all
contrasts (at the second level) of contrasts (at the first level) can
be reformulated as a single contrast at the first level and entered as
such into a simple one sample t-test using the basic stats options.
All the very best - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|