Hi,
If this first-level design and contrasts seems to work ok, then the
second-level (cross-subject) design is trivial - for each of the first-
level contrasts you just run a single-EV group-mean model at second-
level. The only complication is that you can't pass up Fstats in FEAT
at the moment....so you would need to look for positive and negative
group mean responses for the interactions separately.
Cheers.
On 16 Jan 2009, at 14:51, Jan Derrfuss wrote:
> Hi Steve,
>
> Thanks a lot for your reply. I grew interested in what I would get
> if I
> would model the data differently and therefore adapted my EVs. Now I
> have:
>
> EV1 = A on/off
> EV2 = B on/off
> EV3 = C on/off
> EV4 = interaction AB
> EV5 = interaction AC
>
> I ran the 1st-level analysis with the following contrasts:
>
> F F F F F
> 1 0 0 0 0 x
> 0 1 0 0 0 x
> 0 0 1 0 0 x
> 0 0 0 1 0 x
> 0 0 0 0 1 x
>
> My problem right now is that I'm not really sure how to set up the
> 2nd-level
> analysis. It would be really great if you - or somebody else - could
> point
> me in the right direction.
>
> Cheers,
> Jan
>
>
>> Hi - what you're suggesting might be ok - maybe some extra insight
>> can
>> be found in our related example in the manual at:
>> http://fsl.fmrib.ox.ac.uk/fsl/feat5/detail.html#ANOVA3factors2levels
>> Or you might find it easier not to think of this as an ANOVA but as a
>> GLM, modelling the different combination of conditions (which I
>> suspect is actually very close to the way that you are modelling this
>> in your email).
>>
>> Cheers.
>>
>> On 14 Jan 2009, at 13:43, Jan Derrfuss wrote:
>>
>>> Dear All,
>>>
>>> I am currently analyzing a slightly unusual design which I'm not
>>> completely
>>> sure how to set up in FSL. The design is an incomplete 2x2x2-
>>> factorial
>>> design. The design was blocked with all 6 conditions performed
>>> within one
>>> run in counterbalanced order.
>>>
>>> What I've done is the following: There are 6 EVs per subject,
>>> specifying the
>>> onsets of the 6 conditions. The EVs represent the following factor-
>>> level
>>> combinations:
>>> EV1 = A1B1C1
>>> EV2 = A2B1C1
>>> EV3 = A1B1C2
>>> EV4 = A2B1C2
>>> EV5 = A1B2C1
>>> EV6 = A2B2C1
>>>
>>> I set up the following contrasts (for brevity, I skip the reverse
>>> contrasts):
>>> A1 vs. A2: 1 -1 1 -1 1 -1
>>> B1 vs. B2: 1 1 0 0 -1 -1
>>> C1 vs. C2: 1 1 -1 -1 0 0
>>> Interaction AxB: 1 -1 0 0 -1 1
>>> Interaction AxC: 1 -1 -1 1 0 0
>>> (interactions display areas with overadditive activation)
>>>
>>> My questions are:
>>> - Is it possible to set up the analysis this way?
>>> - I noticed that the 2007 course slides suggest a different way to
>>> set up a
>>> 2x2-factorial design (EV1 models A1, EV2 models B1, EV3 models the
>>> interaction, F tests are conducted). Would this be a better way or
>>> an
>>> alternative way? If this approach were to be preferred, how would I
>>> have to
>>> adapt it to my design?
>>>
>>> Any help would be greatly appreciated.
>>>
>>> Cheers,
>>> Jan
>>>
>>
>>
>> ---------------------------------------------------------------------------
>> 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
>> ---------------------------------------------------------------------------
>
---------------------------------------------------------------------------
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
---------------------------------------------------------------------------
|