Thanks Jasper,
So if I understand correctly, I should have only one group ?
But should I create one or two EV for distinguishing patients from controls in a design like:
Design Matrix
Group EV1 Patients EV2 Controls Age Sex
1 0 1 -22 -1
1 1 0 11 1
1 1 0 31 1
1 0 1 -20 -1
Is this more correct ?
Thanks ?
Josselin
Dear Josselin,
> I am performing a statistical comparison of FA maps from two groups (patients and controls), with age and sex as covariates, after having done TBSS. I do not assume an interaction between age x group (neither sex x group).
> I have strange results when changing the design matrices and the use of -D option.
> I have done two analyses:
>
> 1) Design Matrix
> Group EV1(Patients/Controls) Age Sex
> 1 -1 -22 -1
> 1 1 11 1
> 1 1 31 1
> 1 -1 -20 -1
>
> Contrasts
> 1 0 0 (patients > controls)
> -1 0 0 (controls > patients)
>
> With these matrices, I ran the randomize with the -D option (as I read from the forums)
>
>
> 2) Design Matrix
> Group EV1(Patients) EV2(Controls) Age Sex
> 1 0 1 -22 -1
> 1 1 0 11 1
> 2 1 0 31 1
> 2 0 1 -20 -1
>
> Contrasts
> 1 -1 0 0 (patients > controls)
> -1 1 0 0 (controls > patients)
>
> And in this case, following some recommendations on the forum, I did not use the -D option.
I am guessing it comes from your use of "Group" in the second design. Note that, slightly confusingly, "Group" is used for completely different purposes in regular FEAT and in randomise. In randomise it establishes which labels/datapoints that can and cannot be exchanged. It is mainly used for paired designs where one is only allowing exchanges/swaps withins subjects (so as to retain the paired design for all permutations). I am further guessing that you have made a mistake in your example and that your "Group" follows patients/controls? If that is the case it means that every single one of your permutations looks at the original design.
Good luck Jesper
|