Dear Steven,
It doesn't matter too much about the values in the demeaned EVs.
You just need a set that explains the group dependence (of scanner
in your case).
As the data is also demeaned (in either FEAT first-level analysis or
in FIRST vertex-analysis, but *not* in FEAT higher-level analysis) then
you just need two EVs to account for the difference (a third becomes
redundant and therefore rank deficient). There are various options,
all of which are equivalent, and some examples of these are:
1 1 .. 1 0 0 .. 0 0 0 ... 0
0 0 .. 0 1 1 .. 1 0 0 ... 0
OR
1 1 .. 1 -1 -1 .. -1 0 0 .. 0
1 1 .. 1 0 0 .. 0 -1 -1 .. -1
OR
1 1 .. 1 1 1 .. 1 0 0 .. 0
1 1 .. 1 0 0 .. 0 1 1 .. 1
OR
2 2 .. 2 -1 -1 .. -1 -1 -1 .. -1
-1 -1 .. -1 2 2 .. 2 -1 -1 .. -1
(based on three sets: A A .. A B B .. B C C .. C)
These will all perform *exactly* the same job, and after
demeaning none of them will contain zeros.
I'm not sure if one of these corresponds to what you were talking about
or not. For either FIRST vertex-analysis or FIRST first-level analysis then
*any* of these sets will fully model the dependence on scanner location
(or any group: A, B and C).
Hope this helps.
All the best,
Mark
P.S. If there were *no* demeaning then the simplest and easiest to
understand option would be the first two EVs plus 0 0 .. 0 0 0 .. 0 1 1 .. 1.
As in this case you would need three EVs, not two, but the demeaning
always makes the third one redundant, causing rank deficiency.
On 5 Oct 2011, at 09:38, Steven van der Werff wrote:
> Dear Reader,
>
> Currently I am running a vertex analysis. Because my participants were scanned in three different centers I want to include scanlocation as a covariate in my design. When I am using FEAT I demean the covariates and add two EV's. The first has the demeaned scores for scanlocation 1 and 2 with scanlocation 3 defined as 0. The second has the demeaned scores of scanlocation 1 and 3 with scanlocation 2 defined as 0.
> However FIRST demeans on its own so if I assign a value of 0, this 0 will get demeaned and won't be a 0 anymore. So, basically my question is: How do I include a categorical variable as covariate in my vertex analysis design?
>
> Kind regards,
>
> Steven van der Werff
>
|