Hi, this isn't quite what I would recommend; I would just keep the full paired ttest design, and then add a new EV which is like the [ 1 -1 1 -1.... EV, but with each pair of 1 -1 multiplied by the subject's score. Before doing this the full set of scores should be demeaned. Then you can ask separate contrasts about the mean pre-post effect and the correlation of this with score. Cheers
--------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington,
Oxford. OX3 9 DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask]
http://www.fmrib.ox.ac.uk/~steve
----------------------
On 20 Oct 2010, at 20:14, David Paulsen <[log in to unmask]> wrote:
> I know that someone else has just posted a question on this topic, but I'm hoping that I can get an answer related to my format. I've used the design from the FEAT in detail page showing that EV1 contains the information for pre or post (-1 or 1), and the rest of the EVs used to account for within subject variance, one EV per subject. I want to know whether the difference between pre and post is modulated by a covariate score I have for each subject taken at pre. My thought is to modify EV1 or create a new EV with the demeaned measures entered as pre (replacing all the -1's with my covariate), and retain all the pst entries as ones.
>
> Standard
> EV1 EV2 EV3 EV4
> -1 1 0 0
> -1 0 1 0
> -1 0 0 1
> 1 1 0 0
> 1 0 1 0
> 1 1 0 1
>
> Contrasts EV1
> pst-pre 1
> pre-pst -1
>
>
> Covariate
> EV1 EV2 EV3 EV4
> -1 1 0 0
> -1 0 1 0
> -1 0 0 1
> -.3 1 0 0
> .1 0 1 0
> .2 1 0 1
>
> Contrasts EV1
> pst-pre_w_cov 1
> pre-pst_w_cov -1
>
> Does this sound right?
>
> Thank you,
> David
>
|