Thanks a lot for the explanation, Eugene!
Martin
On Tuesday 17 November 2009 16:32:17 Eugene Duff wrote:
> Hi Martin -
>
> > > > 2. In the higher-level group analysis, the differences in event order
> >
> > and
> >
> > > > different event timing for each subject are not an issue since
> > > > first-level variances are carried up from the first-level analysis,
> > > > but
> >
> > I
> >
> > > > presume that I
> > > > should use cope images and not feat directories, because the model
> > > > will be different for each subject.
> > >
> > > The differences in event order and timing should not be an issue. You
> >
> > can
> >
> > > still use Feat directories as you will still have the same number and
> > > order of evs/copes in each lower analysis.
> >
> > Here I have a misunderstanding. The number of EVs will indeed be the
> > same, but
> > if I shuffle the order of events belonging to each contrast, the
> > HRF/model for
> > each contrast will look different for each subject.
> >
> > The documentation on higher-level analysis (
> > http://www.fmrib.ox.ac.uk/fsl/feat5/detail.html ) says:
> >
> > "You can choose whether your higher-level design matrix will be applied
> > to a
> > set of lower-level cope images or a set of lower-level FEAT directories.
> > In the latter, more normal, case, each contrast in the lower-level FEAT
> > directories will have the higher-level model applied..."
> >
> > which doesn't seem to be my case, if I undersdand things correctly?
>
> I can see the confusion. The higher level model is largely independent of
> the the lower level model - it simply models, for example, the differences
> between one group of subjects and another, or, most simply, the mean group
> effect. In a single higher-level analysis you are modelling the cross
> subject (or session, etc) variations in a single cope that you defined in
> all of the first level analyses (e.g. "task1" or "task1-task2"). When you
> select a FEAT directory, FEAT performs separate analyses on all the
> different copes you have defined in the first level models, using the same
> higher level model.
>
> Hope that is clear
>
> Eugene
>
> > Thanks,
> >
> > Martin
--
**********************************
Senior Clinical Research Associate
MRI Unit of the Department of Radiology
Erasme Hospital
Lennik Street 808
B-1070 Brussels
BELGIUM
tel: +32-2-555-4325
fax: +32-2-555-3994
email: [log in to unmask]
**********************************
-----------------------------------------------
Find a way, or make one!
|