Dear Stephen,
thank you very much for the fast response. Probably, the difference
contrast is not what I want.
I would like to see an *interaction* between Factor1 *group* (group1,
grpoup2), and Factor2 *parameter*.
If I split Factor2 (parameter) into two groups (bad performance, good
performance), then its easy to compute the interaction with the contrast
"1 -1 -1 1".
However, I would like to compute the interaction by using the Factor2 in
a parametric way without splitting it into 2 groups.
How I could do such an analysis?
Thank you very much for your advise.
With best regards,
Karsten
Stephen Smith wrote:
> Hi
>
> On 13 Apr 2010, at 10:27, Karsten wrote:
>
>> Dear Colleagues,
>>
>> I have a question concerning the test of parameters in a second level
>> design with two groups. I am using one parameter (lets say
>> "performance") for each subject:
>>
>> group Par group1 Par group 2
>>
>> 1 1.2 0
>> 1 3.4 0
>> 1 -2.5 0
>> 1 4.3 0
>> ... ... ...
>> 2 0 3.3
>> 2 0 1.8
>> 2 0 -2.2
>> 2 0 3.6
>> ... ... ...
>>
>> Both parameters are normalized (mean 0, var 1).
>>
>> To investigate the *negative* correlation between this parameter and
>> the contrast images in two groups I've used the contrast vector '-1 0'
>> for group 1 and '0 -1' for group 2. This revealed a strong *negative*
>> correlation in group 1, whereas the correlation for group 2 was not
>> significant (neither negative nor positive).
>>
>> As a next step I would like to compare both groups. I've used the
>> contrast '1 -1' which revealed highly significant results. But how can
>> this be explained?
>>
>> Does this mean that the general *dependence* between parameter and
>> contrast value is bigger in group1 than in group2, no matter if it is
>> a negative or positive correlation?
>>
>> Or does "1 -1" mean that the *positive* dependence in group1 is bigger
>> than in group2?
>
> Neither - contrast [1 -1] looks for where the correlation with the
> regressor in group 1 is greater than in group 2.
> For example 2>1, 1>0, 1>-1 or -1>-2.
>
> So it seems that the above results are 'contradictory' - which I'm
> guessing therefore just means that the different areas showing up in the
> different contrasts are in different parts of the brain?
>
> Cheers.
>
>
>
>
>> If yes, this would not fit with the results found for the groups
>> separately. Or does this contrast compare the slopes of the
>> correlation, and just show that the one slope is steeper than the other?
>>
>> Thanks a lot four your help,
>> Karsten
>>
>> (The design matrix was constructed with "Glm" and the statistics was
>> computed with "randomise".)
>>
>
>
> ---------------------------------------------------------------------------
> 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] <mailto:[log in to unmask]>
> http://www.fmrib.ox.ac.uk/~steve
> ---------------------------------------------------------------------------
>
>
>
|