When I added columns 3 and 4 in Matlab, the resulting variable didn't
have a mean of zero.
Yes, I was suggesting the use of separate design.mat files, with each
model having 4 total columns (2 columns to model the group mean, and 2
for the separate slopes for each score). If you have all the scores in
the model at once, then it will be difficult to interpret the betas for
individual scores (or the slope difference) given the correlation among
the scores (it appears that you have some scores correlated at r > 0.7).
This is because the contrasts are testing for unique variance in the
context of the other variables in the model
cheers,
-MH
On Wed, 2011-09-07 at 17:16 +0100, Carlos Faraco wrote:
> Michael,
>
> I did mean center the scores, but across all individuals. This was following the advice on Mumford's page.
>
> Do you mean to use one set of scores at a time by having separate mat files for each index score, or do you mean only use contrasts like 10 or 0 -1? In the latter case I would just say that for controls, for example, IM scores are positively correlated with FA on X tract and that for patients they are negatively correlated with X tract, but I would never say anything relating the relationship of the groups to each other?
>
> -Carlos
|