Hi,
On 27 Mar 2009, at 21:25, Christopher Bell wrote:
> Listers,
>
> I am running a TBSS analysis using randomise. I have set up my
> design matrix
> and demeaned my single covariate. From reading the list, I am a bit
> confused
> about modeling the mean, I believe I have done so with EV1 EV2 and
> EV3, and
> hence do not need -D. I have an uneven numbers of subjects in my
> groups, but
> all other values are 0s, so I think this successfully models the
> mean, correct?
Yes - assuming EVs 1-3 are modelling the different group means with 0s
and 1s (looks like it is from below) then that's all correct.
> My other question is about creating the null distribution. I want to
> test
> the differences between three groups of probands, with fairly limited
> subjects in each group: 11,12,14. So comparing each group that
> leaves me
> with 6 contrasts, 7 if I include my covariate. I also will run an F-
> Test on
> two of the independent contrasts (all three groups included).
> Finally to my
> question, can including extra control subjects in my design matrix
> say EV4
> or EV4+EV5 help create a more accurate null distribution, even if I
> end up
> not testing against these EVs at all in my contrast file? Thanks for
> any help.
If an EV doesn't get involved in a given contrast then I believe that
it is effectively ignored wrt the null distribution for that contrast.
Cheers.
>
> EV1 EV2 EV3 EV4
> 0 1 0 (demeaned covariate)
> 1 0 0
> 0 1 0
> 0 0 1
> 0 0 1
> 0 0 1
>
---------------------------------------------------------------------------
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
---------------------------------------------------------------------------
|