Hi Steven
you should take a look onto your design matrix and you will see that they look differently.
'White' is that what counts, i.e. in your first example, where you started with 'A' and proceeded to 'B', (or started with 'zero' and proceeded to 'ones'), the first column of the design matrix should start with 'white' and shifts to grey in the lower half, and for the second column the other way around. So, the first column codes for that group, you have entered first, in this case the 'NORMAL' group.
Therefore, the contrast [1 -1] shows you Normal > Disease, and the contrast [-1 1] shows you Disease > Normals
When you enter the groups as 'BBBBBBBBBBAAAAAAAAA' may guess - without having tested it - is that the first column is grey in the upper half and white in the lower half. Correct?
Here, the first column (and therefore the first value in your contrast) is related to the DISEASE group, since you entered first the scans from the NORMAL group and then those from the DISEASEs group. Therefore are the meaning of the contrasts reversed, i.e. now the contrast [1 -1] tests for Disease > Normal.
Karsten
> I am trying to compare activations across two groups and am getting inverse
> results depending on how the groups are designated in SPM2. For example, if
> my groups consist of 10 images from NORMAL and 10 images from DISEASE, and I
> select the 20 images in the order of N1..N10..D1...D10. SPM2 asks for the
> groups.
>
> SCENARIO WITH LETTERS
> If I designate the groups as letters such as
> A A A A A A A A A A B B B B B B B B B B
> and then choose 1 -1, I assume I am selecting Normal < Disease?
> However, if I designate the groups as
> B B B B B B B B B B A A A A A A A A A A
> and then choose 1 -1, I assumed I was selecting Normal < Disease, but this
> gives me the results of -1 1 from above.
>
>
>
> SCENARIO WITH NUMBERS
> If I designate the groups as numbers such as
> zeros(1,10) ones(1,10)
> and then choose 1 -1 I assume I am selecting Normal < Disease
> However, if I designate the groups as
> ones(1,10) zeros(1,10)
> and then choose 1 -1, I assumed I was selecting Normal < Disease, but this
> gives me the results of result of -1 1 from above.
>
> Is there a "preferred" manner in which to designate groups. If so what is
> recommended? Finally, which input from above gives me normal > diseased?
>
> Thank you very much.
> Steve
>
--
-------------------------------------------------------------
Karsten Specht, PhD
Department of Biological and Medical Psychology
& National Competence Centre for functional MRI
University of Bergen
Jonas Lies vei 91
5009 Bergen
Norway
Tel.: +47-555-86279
Fax: +47-555-89872
[log in to unmask]
http://fmri.uib.no/
|