Dear FSLers,
I have a really simple question but this is very confusing for me.
We have a simple statistical model in TBSS with two groups (controls and patients) and several covariates (e.g. IQ).
So, we did this type of matrix with EV1 (controls), EV2 (patients) and EV3 (IQ):
0 1 90
0 1 72
0 1 85
....
1 0 86
1 0 92
1 0 93
We wanted to look at the difference between groups and also at the effect of the IQ. So we did these contrasts:
1 -1 0 pat>contr
-1 1 0 contr>pat
0 0 1 IQ+
0 0 -1 IQ-
The two first contrasts looked fine but for the last two, [0 0 1] all the skeleton is almost at 1 and [0 0 -1] everything is alsmot at 0.
Do I have to change of contrast or to add a constant somewhere or to demean EV3 or to do only one EV for the group effect (-1 1)?
Thanks for your help.
Hervé
|