Hi,
We recommend that you demean your scores as a single set of values before splitting them into separate groups (for your separate EVs). If you do this then you still get the adjustment for any consistent difference between the mean scores in the two groups.
The demeaning method won't actually affect the contrasts that only look at the correlations of the scores, but will affect the contrasts that look at the mean activations of the two groups. However, your contrasts are not right. If your first two EVs are
indicators (values of 1 for one group and values of 0 for the other group) and the second two EVs are the demeaned performance scores, then the contrasts you would use would be:
1 0 0 0 mean patient activation
0 1 0 0 mean control activation
1 -1 0 0 patients > controls
-1 1 0 0 controls > patients
0 0 1 0 patient (positive) correlation - of performance score and BOLD activation
0 0 -1 0 patient (negative) correlation
0 0 0 1 control (postive) correlation
0 0 0 -1 control (negative) correlation
0 0 1 -1 patients correlation > controls correlation
0 0 -1 1 controls correlation > patients correlation
I hope this helps.
All the best,
Mark
Dear FSL users,
I am fairly new to using FSL feat and wanted to make sure that my contrasts were set up correctly for my analysis. We have two groups (patients and controls) who performed the same task. I want to look at which brain regions that are activated while performing
the task in each group. The EVs are listed in 2 separate groups because we expect the variance to differ. So I have set up the group flame as:
patients > controls 1 -1
patients < controls -1 1
Additionally, I want to see if there is an effect of task performance within each group. I have entered in the scores as a separate EV. Should I demean the data and if so, is it within each group separately if I am only interested in the correlations within
group?
I have set up a preliminary contrast in this way (after demeaning the score EV within each group):
|
patient |
control |
score patient |
score control |
| patients correlation |
1 |
0 |
1 |
0 |
| controls correlation |
0 |
1 |
0 |
1 |
| mean score patients |
0 |
0 |
1 |
0 |
| mean score controls |
0 |
0 |
0 |
1 |
Thanks,
Corinna
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
