Dear Vitaly,
> I am analysing SPECT images of Lyme Disease subjects and I seem to be
> running into "illegal contrast" problem. My design is as follows:
>
> Lyme Patient Control Group Design type:
> --------------- ---------------
> Rest Challenge Rest Challenge Multistudy, different conditions
> ---------------------------------
> 1 0 -1 0
> -1 0 1 0
> 0 1 0 -1
> 0 -1 0 1
> ---------------------------------
>
> The objective is to compare individual Lyme patients against a group of
> controls with the presence and the absence of Diamox. The program runs to
> completion, but the output is very strange. The matrix of contrasts on the
> SPM{F} page looks like:
>
> 0.5 -0.5 -0.5 0.5
> -0.5 0.5 0.5 -0.5
> -0.5 0.5 0.5 -0.5
> 0.5 -0.5 -0.5 0.5
>
>
> What is interesting, a similar analysis in SPM'95 works fine.
>
> Are these contrasts illegal? If so, what could be done to implement such a
> design?
It looks as if the contrasts have mean corrected within group. This
may be something that Andrew has included to ensure estimable
contrasts. I think perhaps he better comment on this. The contrasts
actually used by SPM (the second set) are all interactions and are
valid. These contrasts are, I think, what you want. To look at the
simple main effects (your first set of contrasts) you should perform
separate analyses under Rest and Challenge.
All the best - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|