> What I'm not clear on is the creation of the con*.img files. Let's say I
have
> six groups of subjects with 10 subjects in each group, and each subject
> received a scan which alternated condition A with baseline. I'm
interested in
> the effect of condition A. A single subject's analysis is easy, and
creates a
> con*.img file. Do I have to specify a fixed-effects design matrix (with
> global scaling) that includes all 60 subjects at the same time, and then
go
> through it one subject at a time getting a contrast for A vs. baseline for
> each subject, which I can then use in the second-level analysis; or can I
use
> each subject's con*.img from their own data analyzed individually?
The con*imgs created by the two methods you describe will be identical
(assuming you modelled the subjects as separate sessions in the big
fixed-
effects model, ie the covariates are separable), so if you have done the
former single-subject analyses already, then you might as well use those
con*imgs (and save on swap space too ;-)
> Also, can a random-effects design matrix be set up in SPM99b under the
fMRI
> section, or do we have to use the PET/SPECT models? We are trying to set
up
> the second-level design matrix before we have the data, and the basic
models
> in the fMRI section require that we have the data already. (We're doing
this
> all through the GUI.)
For basis statistics (eg two-sample t-tests) you don't need to switch to
PET.
For more advanced models, you will need to (eg for six groups).
Best wishes
Rik
---------------------------8-{)}-------------------------
DR R HENSON EMAIL [log in to unmask]
Wellcome Department of
Cognitive Neurology TEL (work1) +44 171 833 7483
12 Queen Square TEL (work2) +44 171 833 7472
London, WC1N 3BG FAX +44 171 813 1420
URL: http://www.psychol.ucl.ac.uk/rik.henson/index.html
---------------------------------------------------------
|