Jack,
A paired t-test is what you need. You can do this in the way your
colleague suggested:
> Also, someone else (an SPM user) suggested that I should do the A-B contrast
although not as a contrast, because as you have separate runs for A and B,
you would do your first-level analyses on all of your A and B runs data
separately and then do a straight subtraction between the A copes and B
copes (using avwmaths for example) and then pass the A-B subtractions up
to a group analysis to look for the group mean of A-B.
Alternatively, rather than using avwmaths you can do it all in FEAT using
a second level design matrix in the way suggested in the paired t-test
example at:
http://www.fmrib.ox.ac.uk/fsl/feat5/index.html
Your other issue is at the second-level. Because you only have 3 subjects
you can't model a separate between-subject RE variance and hence you
should lump the subjects and sessions together in the second-level.
Cheers,
Mark.
Mark Woolrich.
Oxford University Centre for Functional MRI of the Brain (FMRIB),
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK.
Tel: (+44)1865-222782 Homepage: http://www.fmrib.ox.ac.uk/~woolrich
On Mon, 22 Mar 2004, Jack Grinband wrote:
> Hi All,
> I'm a bit confused about how to set up my group analysis. I have 3 subjects with 5 sessions each.
> Each session had 2 runs, A and B. I want to do a contrast between A and B.
>
> My intuition is to set up a second level analysis for condition A with an EV for each subject (i.e. 1
> group, 3 EVs). Then a separate second level analysis for condition B. Then I'd do a third level
> analysis to compare the group A activation versus group B activation.
>
> However, the second level analysis generates a .gfeat directory, but the group analysis requests a
> .feat directory. It's not clear to me what the input to my third level analysis should be (assuming
> my intuition was correct in the first place).
>
> Also, someone else (an SPM user) suggested that I should do the A-B contrast as a first level
> analysis and bring the contrast up to the second level analysis for the group. Doesn't this assume
> equal variance across A and B?
>
> thanks,
>
> jack
>
|