Dear Angela,
That design matrix looks fine. Whether you split the covariates into separate EVs for each group or not is a question of what you are trying to test and has been covered in quite a number of emails on the list and also in Jeanette Mumford's great webpages on the subject. However, it is perfectly reasonable to do exactly what you've done.
I do not know why you are getting the error and we need to see the exact command line from randomise before we can help further.
All the best,
Mark
On 10 Dec 2012, at 14:20, Angela Favaro <[log in to unmask]> wrote:
> Dear FSL experts,
> I am getting crazy with a matrix I have prepared with glm_gui to be used with TBSS (41 patients and 18 controls).
>
> It is a simple matrix with two groups and two covariates of no interest.
> Contrasts: 1 -1 0 0 and -1 1 0 0
>
> EVs:
>
> group EV1(patients) EV2 (controls) age (dem) lat (demeaned)
> 1 1 0 13.9 -1.64
> 1 1 0 -12.5 -1.64
> 1 1 0 12.9 0.36
> 1 1 0 3.34 0.36
> 2 0 1 24.4 0.36
> 2 0 1 20.8 0.36
> 2 0 1 -1.7 0.36
>
> age and lateralization were demeaned across groups and put in one column20 since I am not interested in correlations
> However I have the same problem splitting the age and lat EVs in two
> I have used similar matrices in other studies without problems, but now the software says:
>
> Problem with processing the model: child killed: segmentation violation
>
> What am I doing wrong?
> Thank you!
>
> Angela
>
|