Hi.
On Tue, 14 Jan 2003, Martin Kronbichler wrote:
> thanks for your fast response. So the problem with different TRs has
> nothing to do with unbalanced designs. Am i right in assuming that scaling
> all scans from all subjects to a commom mean is the critical point, so if
> scaling is used any mixed (random) effects model would be valid?
Yep, that makes sense.
> So i couldnīt analyse the data from all subjects in one first level model
> (fixed effects model, not that i want to do that)but in any second level
> (mixed effects)?
not quite sure what the question is here - having used grand mean
intensity normalisation (scaling), both second-level fixed-effects and
mixed-effects analyses should be fine.
> Just a second short question, i have talked to the people who did this
> study and it contains two groups, controls and patients (parkison disease)
> and the TR seems to be dependent on brain volume in this particular
> sequence. Could i run into any problem if the TRs of the both groups are
> significantly different (which i donīt know, but will check), because if
> scaling is the solution, differences in TR beetween groups shouldnīt be any
> problem, i assume?
indeed - hopefully, if my previous answer is correct (and a physicist here
assures me it should be so!) then there should be no problem here. you
may well end up with different variances in the different groups, so the
ability to model the two group variances separately in FEAT should be of
value here.
Thanks, Steve.
>
> Thank you very much,
> Martin
>
>
> Martin Kronbichler
> Institute of Psychology
> University of Salzuburg
> Hellbrunnerstr. 34
> 5020 Salzburg
> Austria
> tel.: 0662/8044-5162
> fax.: 0662/8044-5126
> e-mail: [log in to unmask]
>
>
>
> >Hi - yes, FEAT5 does indeed allow you to analyze unbalanced designs, but
> >in fact, that's not really the point here:
> >
> >The first thing to note is that changing TR from 0.2 to 0.6s probably
> >won't have a huge impact anyway on the baseline and activation BOLD
> >signal. However, in any case, any effect would largely be a scaling of
> >both baseline and activation signal change, so the
> >percentage-change-activation would be the same, ie the grand-mean (4D)
> >intensity normalisation that FEAT always carries out in the preprocessing
> >would rescale the baseline AND the activation signal so that it is
> >comparable across subjects.
> >
> >So - you should hopefully be safe in your analysis.
> >
> >Thanks, Steve.
> >
> >
> >
> >
> >
> >
> > Stephen M. Smith
> > Head of Image Analysis, FMRIB
> >
> > Oxford University Centre for Functional MRI of the Brain
> > John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
> > +44 (0) 1865 222726 (fax 222717)
> >
> > [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>
Stephen M. Smith
Head of Image Analysis, FMRIB
Oxford University Centre for Functional MRI of the Brain
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
|