Splitting your "on" sections into two halves is a rather unique design.
Is the hypothesis that the signal is decaying over the course of the
"on" block? If so, it might be better to model that directly with a
"sustained" EV and a linearly decaying EV...
In your current model, the (1,-1) contrast between the first and second
half EVs should have been included as a contrast in your 1ST LEVEL
analyses, so that you can properly account for the correlated nature of
those two EVs.
cheers,
-MH
On Fri, 2011-09-09 at 09:52 -0400, Leslie Engineering wrote:
> To clarify there was an "on" "off" stimulation paradigm. I divided the
> "on" sections in half to see response difference in first and second
> half of the stimulation.
>
>
> I analyzed each individual subject using text files representing the
> first and second half paradigms and used contrasts (0,1), (1,0), and
> (1,1) to see an overall stimulation activation as well as the halves
> (EV1 and EV2).
>
>
> Now I want to look at an overall group picture of activation
> differences in first and second half i.e. EV1>EV2 and EV1<EV2.
>
>
> I initially tried just using EV1 with all ones in the full model
> wizard because I thought the contrasts from individual analysis would
> remain. But, instead I just get one zstat/thres_zstat* file.
>
>
> So next I tried 2 evs in higher level and added the contrasts (1,-1)
> (-1,1) and under the EV panel made two columns of ones. But there is
> an error message.
>
>
> Any advice would be SOOOOOO appreciated!
|