If you are looking for the difference in the correlation between the
groups then you have to use DODS and 0 0 1 -1. The way to understand
it is that you have four regression coefficients and each means
something. The way you have it set up in the fsgd file is that they
are
1. group 1 intercept
2. group 2 intercept
3. group 1 cognition slope
4. group 2 coginition slope
Each coefficient gets multipled by its corresponding weight in the
contrast and then the results are summed. So if you use 0 0 1 -1 you
end up with contrast = g1cogslope-g2cogslope which is the difference
you want
If you are just interested in whether the cog slope changes without
regard to the group, then you would use 0 0 0.5 0.5 which means
contrast = 0.5*g1cogslope + 0.5*g2cogslope which is the average of
the slopes. In this case, using DOSS and 0 0 1 will give you the
same results if there is no difference in the cog slopes between the
groups
On 1/12/2023 8:48 AM, Agurne Sampedro
Calvete wrote:
External Email - Use
Caution
Dear FreeSurfer experts,
I have two groups and two time points and I would like to
do a correlation analysis between a variable (cognition) and
the change score in cortical thickness between the two time
points (time2 minus time1).
I would like to see if the correlation is different
among each group, if there is an interaction with the
group.
The paired-diff.fsgd file (second fsgd file) is the
following:
GroupDescriptorFile 1
Title OSGM
Class Group1
Class Group2
Variables cognition
Input subject1pair Group1 0.546
Input subject1pair Group2 -0.243
...
How should it be the contrasts files to see the correlation
of the variable cognition with the change score in
thickness taking into account the groups? And should I use
DODS or DOSS?
Using DODS:
0 0 0.5 0.5
or
0 0 1 -1
I don't understand the difference between both contrasts. I
think I should use the one of "0 0 1 -1" to see whether the
cognition variable correlates with change in thickness
differently according to the group. Is this correct?
