Hi,
On 4 Apr 2008, at 02:55, Liberty Hamilton wrote:
> Dear FSLers,
>
> I am writing to confirm the design matrix I have been using to
> investigate
> changes in activation during a block design task, as I have not been
> able to
> find anything specifically on FSL's web site about my particular
> design, and
> I am a bit confused as to how I should proceed, and I haven't
> managed to
> find anything specifically about this problem on the email list.
> The answer
> should be quite simple, but I want to make sure that I'm doing
> things correctly.
>
> The problem is as follows. I am performing a higher level analysis
> where
> there are two groups (patients and controls) with about 20 subjects
> in the
> patient group and 35 subjects in the control group. The subjects do
> not
> differ significantly in age nor in sex distribution, but I would
> like to
> include sex and age as covariates to minimize effects that might be
> accounted for by these variables. So, at the moment I am creating a
> design
> matrix that looks as follows:
>
> Group control patient age sex
> 1 1 0 21 1
> 1 1 0 24 -1
> 1 1 0 35 -1
> 1 1 0 42 1
> 1 0 1 23 1
> 1 0 1 43 -1
> 1 0 1 22 1
That looks fine - you might want to remove even more variance by
having separate age and sex EVs for each of the subject groups.
> I am also orthogonalizing the age and sex variables with respect to
> EV1 and
> EV2 (controls and patients). The groups are all assigned to one
> since I
> cannot estimate separate group variances when covariates are added
> to my
> model. So then, in my contrasts, I would like to look at
> differences in
> activation between patients and controls, and I do this:
>
> Contrast control patient age sex
> ctrl>pat 1 -1 0 0
> pat>ctrl -1 1 0 0
> avg ctrl 1 0 0 0
> avg pat 0 1 0 0
>
> with no F tests. Is this correct? Am I correct in assuming that
> the test
> I'm performing with controls > patients and patients > controls is
> still an
> F test because age and sex are included in the model, or do I have to
> specify these explicitly? If I do the F test, how can I determine
> directionality of the contrast?
No, these 4 contrasts are not f-tests - they are all single,
directional, contrasts. The presence of the other covariates is not
related to whether these contrasts are t-tests of f-tests. An example
of an f-test would be if you did an f-test across your first 2
contrasts, which would then ask is there any difference in either
direction between the groups.
Cheers.
>
>
> If anyone could please confirm that I have gone about this in the
> right (or
> wrong!) way, I would very much appreciate it.
>
> Thank you so much in advance!!
>
> Liberty Hamilton
> Laboratory of Neuro Imaging at UCLA
> http://www.loni.ucla.edu
>
---------------------------------------------------------------------------
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
---------------------------------------------------------------------------
|