HI - yes you can do this in one big matrix but it may be easier to just run two separate analyses.   For the separate groups with this non-paired design you should just have one EV per group, with just 1s and 0s in each.

Cheers.



On 18 Jun 2010, at 01:25, Christopher Bell wrote:

FSLers,

I have four groups. I only really want to compare group 1 vs. group 2 and group 3 vs. group 4 for the time being. Is this appropriate to do in a single design matrix? Basically the comparison is chronic probands vs. older controls and early-onset probands vs. young controls. I tried just comparing group 1 vs group 2 in my first design which had only had two columns and I just left groups 3 and 4 as zeros for both columns, but this generated the error below. My new 4-EV design is proposed below, but I have not tested this yet. I would then test these 4 contrasts. The reason I have one file, is next I will likely combine the chronic/early-onset subjects together as a single group using age as a covariate. Thanks for any comments.

EV1 EV2 EV3 EV4
1   -1   0   0
-1   1   0   0
0    0   1   -1
0    0   -1   1

1 0 0 0 1 0 0 0
1 0 0 0 0 1 0 0
0 1 0 0
0 1 0 0
0 0 1 0
0 0 1 0
0 0 1 0 0 0 0 1 0 0 0 1

Warning-design matrix uses different groups (for different variances),
but these do not contain "separable" Evs for the different groups (it
is necessary that, for each EV, only one of the groups has non-zero
values).

Chris



---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director,  Oxford University FMRIB Centre

FMRIB, JR Hospital, Headington, Oxford  OX3 9DU, UK
+44 (0) 1865 222726  (fax 222717)
[log in to unmask]    http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------