The easiest way of doing this would be a two-sample t-test with
covariate. The two-sample t-test gives you the group membership coding,
while the covariate will model your behavioural score.
Volkmar
Am Dienstag, den 25.11.2008, 08:38 +0000 schrieb cyril pernet:
> I there,
>
> ok you were not far .. ;-)
>
> you need the multiple regression model witout cste (prob access via
> spm_spm_ui rather than the GUI since I never managed to have no cste
> otherwise .. )
>
> 1st reg. the group belonging say for 2*3 subjects (hope you have more
> though .. ): 111-1-1-1
> 2nd regressor: behavioural scores group one: X X X 0 0 0
> 3rd regressor: behavioural scores group two: 0 0 0 X X X
> other regressors = confound, e.g. age, and you have to input the total
> grey matter
>
> contrasts:
> 1 0 0 0 0 = group 1 > group 2
> -1 0 0 0 0 = group 1 < group 2
> 0 1 0 0 0 = + corr in group 1
> 0 0 1 0 0 = + corr in group 2
> 0 1 -1 0 0 = corr in group 1 > corr in group 2
> ---> it is important to 1) see what;s happening in each group and 2)
> plot the data --> you may for instance have corr gp1 > corr gp2 because
> gp2 have no corr vs. a smaller corr ..
>
> gd luck
> cyril
>
>
> > I have been attempting to determine the most appropriate design to use for a
> > second-level analysis for determining group differences in the correlation
> > between a behavioral score and the brain.
> >
> > From thoroughly reading the archives as well as available information on
> > SPM5 designs, it seems that choosing design=multiple regression with 3
> > covariates 1) behavioral score, 2) indicator variable for group 1, and 3)
> > indicator variable for group 2 is recommended followed by t-contrasts for
> > group, e.g., 0 1 -1 0 and 0 -1 1 0.
> >
> > The problem with the above model seems to be the inability to look at the
> > "simple effects", i.e., the estimated regions with "significant" positive
> > betas for group 1, estimated areas with "significant" negative betas for
> > group1....etc.
> >
> > I would have thought I could specify 0 1 0 0, 0 0 1 0, 0 -1 0 0, 0 0 -1 0
> > but this simply is not allowed by SPM.
> >
> > At any rate, this seems to be a fairly straightforward analysis question:
> > Are the group differences in the correlation. If there are group
> > differences, what are the within group betas? These with group betas are the
> > only way to describe the reason for the differences.
> >
> >
> > Any help, input, would be greatly appreciated. i realized i may have phrased
> > this question in many ways recently but this one last attempt to get an
> > answer. If i am missing something obvious that everyone should know please
> > point me in the correct direction. I greatly need a push in the correct
> > direction.
> >
> >
> >
>
>
--
Volkmar Glauche
-
Department of Neurology [log in to unmask]
Universitaetsklinikum Freiburg Phone 49(0)761-270-5331
Breisacher Str. 64 Fax 49(0)761-270-5416
79106 Freiburg http://fbi.uniklinik-freiburg.de/
|