Hi,
On 23 Apr 2008, at 20:51, Jeff Spielberg wrote:
> Hi, I have a question about Feat higher-level analysis design
> setup. I have
> one group of subjects who completed two tasks (tasks A and B) along
> with a
> questionnaire measure. I want to correlate the questionnaire scores
> with
> activation for each task and then compare the two beta maps (to find
> significant differences between correlations). I'd like to have a
> design
> that does this simultaneously, instead of having to run an HLA for
> each task
> and then another that compares the Cope outputs. After looking at
> the Feat
> webpage I've come up with a potential design (combining the Paired
> Two-Group
> Difference and Single-Group Average with Additional Covariate
> designs) that
> I want to make sure is correct. I set up an EV for each subject's
> individual mean. I also have one EV that has questionnaire scores
> in each
> input corresponding to task A and the same questionnaire scores
> multiplied
> by -1 in the inputs for task B. I would then set up one Cope with a
> 1 for
> EV1 and zeros everywhere else to get at the comparison I want. I
> left all
> the Group inputs set as 1 to signify that all subjects are from the
> same
> group. Below is an example of the design,
>
> Input Group EV1 EV2 EV3 EV4 EV5 EV6
> subject 1 task A 1 5 1 0 0 0 0
> subject 2 task A 1 3 0 1 0 0 0
> subject 3 task A 1 7 0 0 1 0 0
> subject 4 task A 1 2 0 0 0 1 0
> subject 5 task A 1 8 0 0 0 0 1
> subject 1 task B 1 -5 1 0 0 0 0
> subject 2 task B 1 -3 0 1 0 0 0
> subject 3 task B 1 -7 0 0 1 0 0
> subject 4 task B 1 -2 0 0 0 1 0
> subject 5 task B 1 -8 0 0 0 0 1
>
>
> Is this design correct?
Yes, this is fine, although you might want to separate out the mean
task difference from the mean difference modulated by the
questionnaire score:
EV1 EV2
1 0
1 -2
1 2
1 -3
1 3
-1 0
-1 2
-1 -2
-1 3
-1 -3
where the new EV2 is the old EV1, with demeaning the subject scores
before multiplying by the new EV1. I _think_ this makes sense....
> Also, if I didn't want to model each individual
> subject's mean separately would it be valid to have the same design
> as above
> except one column of 1's instead of EVs 2 though 6 (this seems more
> similar
> to doing two HLA's and comparing the output as a paired t-test)?
That would give you the equivalent of an unpaired t-test and is less
sensitive; I wouldn't do that.
Cheers.
> Thanks for
> your time,
> Jeff
>
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|