Hi,
If all you have is one group of subjects and you want to look for a
correlation
with some other variable then it is perfectly reasonable to do what
you are
doing. It is just more common to have scanned two groups and be looking
for a difference. However, in the case where all you want is to see
the correlation
with one variable, then your design is absolutely fine.
Note that there are some changes in FIRST coming up soon in the next
patch
(days away) which will also substantially improve the sensitivity.
All the best,
Mark
On 19 Mar 2009, at 16:53, Quincy Rondei wrote:
> Hi,
>
> First of all, thank you for all your help.
> In response to the question: I am not sure this is de design I want.
> I am however sure I want to do a comparison based on a covariate. Is
> this at all possible? If so, would removing the first EV (the rows
> of 1's) from my GLM give sensible results?
> Or am I on the wrong track and should I divide the group in two (a
> "hi" scoring and a "low" scoring group)? If so, Should I use 1 EV
> ( with 1 and -1, or 1 and 0 denoting the group membership?) or 2
> EV's with similair numbering?
>
> The GLM I originaly used was similair to the following:
>
> Group EV1 EV2
>
> Mean V1
> Input1 1 1 .125
> Input2 1 1 - 5
> Input3 1 1 3
>
> All the best,
>
> Quincy
|