Hi Stijn,
When thinking about the impact of demeaning, it is very useful to
actually write out your model. Demeaning a continuous variable is
effectively adjusting where the "intercept" of that slope falls along
the axis of the continuous variable. That is, if you have a mean age of
say 20, and use a demeaned variable for age, then the intercept
component of the model is the estimated value at age 20.
It gets confusing, but demeaning a continuous variable really only has
an impact on the inferences typical in neuroimaging when modeling an
interaction, in which case the impact is on the inference of the
CATEGORICAL variable, not the continuous variable. This makes sense
when you think about it because the slope estimate itself is independent
of any constants added to the continuous variable.
Best,
-MH
On Mon, 2011-02-14 at 12:15 +0000, Gwenaƫlle DOUAUD wrote:
> Hi Stijn,
>
> > It seems logical to demean gender, since the "mean gender"
> > is not equal for the groups (groups aren't matched). What do
> > you exactly mean by splitting the age per group? There might
> > be interaction with other variables that are continuous.
>
> Please have a look at the archives:
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1102&L=FSL&P=R20977&1=FSL&9=A&I=-3&J=on&d=No+Match%3BMatch%3BMatches&z=4
>
> > So if I demean the gender column, the design should be
> > okay? How about the randomise script execution? Is this
> > correct:
> > randomise -i all_FA_skeletonised -o tbss -m
> > mean_FA_skeleton_mask -d design.mat -t design.con -n 5000
> > --T2 -V
>
> Looks fine to me...
>
> Cheers,
> Gwenaelle
>
>
> --------------------------------------------------------------------
>
> Gwenaƫlle Douaud, PhD
>
> FMRIB Centre, University of Oxford
> John Radcliffe Hospital, Headington OX3 9DU Oxford UK
>
> Tel: +44 (0) 1865 222 523 Fax: +44 (0) 1865 222 717
>
> www.fmrib.ox.ac.uk/~douaud
>
> --------------------------------------------------------------------
>
>
>
|