Hi again
> The message you directed me to says that the two runs must be
> registered together.
> In the first-level analyses of the two runs, I used the same anatomical
> data set during the registration portion since all scans were collected
> during the same scanning session.
> Are the COPES derived after registering to the anatomical data set or
> before?
The copes after the 1st level will be in the space of their respective
example_func images. If these are in register, then you can go ahead,
otherwise you'll have to apply some transformations.
There are a few different ways to put them in register. I would probably
put them all in mni space
flirt -in stats/COPE -ref reg/standard -applyxfm -init
reg/example_func2standard.mat -o COPE_to_standard
and then do the analysis on the COPE_to_standard
Does this make sense?
> Am I ok in assuming the two functional datasets are registered to each
> other in my case?
I'm not sure - have a look. Are they lined up?
>
> -Brad
>
> p.s. We covered your Nature Neuroscience paper on connectivity in a
> journal club recently. It was well-received. Paul Matthews has visited
> us here in Calgary, and at that time we informally discussed possible
> collaborative efforts in the future. Connectivity is something we
> really want to get into, especially for our research in Multiple
> Sclerosis. We are a relatively new lab, and we're just getting off the
> ground. I've registered for the Neuroinformatics and the Brain
> Connectivity courses in Cuba next month. I look forward to meeting you.
>
Thanks for the interest, I look forward to meeting you too - particularly
in Cuba!!
Cheers
Tim
> On Mar 23, 2004, at 3:38 AM, Tim Behrens wrote:
>
> > Hi Brad -
> > So I think you can use any post-stats settings you want, but yes
> > cluster
> > will be absolutely fine.
> >
> > The main issue here is how to compute the stats. I don't think you
> > really
> > want to be using random/mixed effects analysis to compute the "between
> > run" variances. I think you probably want to analyse the data "as if"
> > you
> > were able to concatinate them.
> >
> > To do this you have to do a fixed effects analysis at the between run
> > level. Unfortunately, we don't at present have a fixed effects option
> > for
> > the GLM. We will put one in the next Feat release.
> >
> > However, if you only have a single EV, you can follow the instructions
> > in
> > (for example)
> >
> > http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0211&L=fsl&P=R3599&I=-1
> >
> >
> > Hope this helps
> >
> > Tim
> >
> > On Mon, 22 Mar 2004, Brad Goodyear wrote:
> >
> >> Hi.
> >>
> >> I'm wanting to analyze a data set comprised of multiple scans within
> >> the same session for the same task.
> >> Instead of concatenating runs (unadvised, I know), I want to perform
> >> the high-level analysis on the low-level analyses to get the mean
> >> activation across all scans, but still want to get the results in a
> >> cluster analysis. What are the correct settings for the post-stats?
> >> Should I be using "Cluster correction" with a chosen Z level (say 2.3
> >> -
> >> as used in the low-level analyses?), then set P to 1.0? setting p to
> >> 0.01 eliminated all activation.
> >>
> >> -Brad
> >>
> >
> > --
> > -----------------------------------------------------------------------
> > --------
> > Tim Behrens
> > Centre for Functional MRI of the Brain
> > The John Radcliffe Hospital
> > Headley Way Oxford OX3 9DU
> > Oxford University
> > Work 01865 222782
> > Mobile 07980 884537
> > -----------------------------------------------------------------------
> > --------
>
|