Hi Virve.
> In my design, there are two types of tasks, L
> and C, and each task has two "situations", L1 and L2, and C1 and C2.
The problem is caused by not having a separate "rest" condition - so any
given condition (say C2) is just the inverse of the sum of the other 3
(that's what is meant by linear combination) - this is known as a rank
deficient design. This is not a disaster, depending on what questions you
want to ask, but it is often conceptually a lot easier to form simple
contrasts (ie ask certain questions) if you also include a rest condition.
(In the case of also having a rest condition, you would then model your 4
EVs and would not explicitly model the rest condition.)
The best thing to do with your data then is to exclude one of the EVs from
the model (for example C2). You then continue by considering C2 as your
"rest" condition. Thus your three EVs are L1, L2, C1.
To ask question L1>L2 use contrast [1 -1 0]
To ask question C1>C2 use contrast [0 0 1]
To ask question L1>C1 use contrast [1 0 -1]
To ask question L2>C2 use contrast [0 1 0]
etc. Note that you cannot ask L1 compared with "true rest" (etc) as you
don't have a true rest condition in the experiment.
Setting the orthogonalisation option will not help solve this problem in
this case.
> I would also be interested to run an ANOVA on my data and study the main
> effects and interactions. If I understood the instructions of the manual
> correctly, to study an interaction, I should make an EV to represent an
> expected interaction (with zeros and ones). So in my design there would be
> several of these interaction EVs. How to avoid the problem presented in
> the beginning of this message?
As these are four different conditions, you can't look for interactions as
the conditions never overlap (in time) with each other - interactions
would be comparing the response when two conditions are happening at the
same time with what happens when they happen separately.
Hope this all makes sense - feel free to ask further questions.
Thanks, Steve.
--
Stephen M. Smith
Head of Image Analysis, FMRIB
Oxford University Centre for Functional MRI of the Brain
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
|