Thomas,
Have a look at a recent reply given to a similar question.
http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0303&L=fsl&D=0&I=-1&P=611
As there would recommend doing a 3-level analysis. Time series
concatonation is troublesome (with motion correction,
temp autocorrelation etc).
Cheers, Mark.
Mark Woolrich.
Oxford University Centre for Functional MRI of the Brain (FMRIB),
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK.
Work: +44-(0)-1865-222713, Mobile: +44-(0)-7808-727745
Email: [log in to unmask]
On Wed, 21 May 2003, Thomas Thesen wrote:
> Hi,
>
> I have data sets with 4 sessions per subject, where the 4 sessions are
> acquired in direct succession and are virtually identical in terms of
> experimental design ( same number of trials/ condition, ISI's etc.).
> However, physically they are 4 different time-series data sets.
> As far as I understand I could
> (i) concatenate these 4 data sets into 1 before doing any preprocessing and
> 1st level analysis, and then proceed with 1 data set / subject.
>
> (ii) do a Multi-Session & Multi-Subject (Repeated Measures - Three Level )
> analysis whereby any preprocessing and 1stlevel analysis is done on the
> individual 4 session data sets first. The results are then fed into one 2nd
> level analysis. The group mean effect for any given condition would then be
> extracted through a 3rd level analysis.
>
> I would appreciate if you could tell me whether both approaches are sound
> and why you would recommend one over the other.
>
> Thanks a lot.
>
> best wishes,
>
> Thomas
>
|