|
Thanks for the kind words, but hardly elegant....
To my recollection the issue with SPM regions was that both spm_regions and spm_peb_ppi were
adjusting for the null space of the contrast. This resulted in slight but not clearly insignificant
changes to the deconvolution. I believe this was fixed by Karl, subsequently.
The problem that Denise asks about below (which has been mentioned on the list previously) pertains
to how SPM deals with multi-session data beyond that of design matrix. spm_graph for example will
only look at 1 session at a time, dcm- only one session, etc. The way around this, for the moment,
has meant analyzing the data as if it all came from one session but also including regressors for
the block effects. Of course this may mean analyzing the data again...
So to do this just put all your sessions together as if they were 1 large session. Adjust the event
onset times accordingly. You may also want to remove event onsets that occur right near the end of
each session, i.e., perhaps within 5-10 seconds or so. Do not remove any of the scans. Then you
would include user defined regressors for the block effects. The number of regressors will be the
number of sessions - 1. For example if you have 3 sessions, you will have 2 block effect columns,
and your design would be
eventOnsets blockEffect-1 blockEffect-2 mean (the mean column is put in by SPM).
If your runs were 5 scans each then the block effect columns would be entered as follows:
1 0
1 0
1 0
1 0
1 0 <- end of first session
0 1
0 1
0 1
0 1
0 1 <- end of second session.
0 0
0 0
0 0
0 0
0 0
The last session's block effects are estimated by the mean column.
This turns out to have a somewhat different design matrix then if you put in the 9 sessions
separately. In that case there would be 9 block effect columns and no mean column. Also the block
effect columns for the 1-session design are convolved with the high-pass filter by SPM. This is not
done in a multi-session design. I've anecdotally looked at studies estimated both ways. In one case
I set up the design as described above for one session. In the other I've set up a single session
design and then munged the design matrix and various SPM.mat fields so the block effects were
exactly those of a multi-session study. The final results differed at the 0.01 level or less for
the t-values and not at all for the locations, but again this is just anecdotal.
Once you have set up your design this way, setting up PPI's, plotting, etc. is easy. Well perhaps
not easy but at least not hindered by the multiple sessions.
Good luck,
Darren.
==============Original message text===============
On Tue, 16 May 2006 10:04:32 am CDT Tobias Egner wrote:
Hi Denise,
your problem is due to PPI not dealing well (in fact, not dealing at
all) with multi-session data. When you set up your PPI, you will have
to do it run by run. When you do this, unfortunately, SPM will include
zero columns for each other run, so when you do a PPI on your first
run (out of 9 in your case), SPM will include 8 columns of zeros in
yX.X0 in the VOI.mat variable, and these zero columns will later cause
your PPI.Y to be full of NANs. One (slightly tedious) way of avoiding
this is to manually remove the zero columns from the VOI.mat file (in
yX.X0).
I think Darren Gitelman has posted a more elegant solution involving
changing some SPM code in SPM_regions...
good luck,
Tobias.
Tobias Egner, PhD
fMRI Research Center
Columbia University
Neurological Institute, Box 108
710 West 168th Street
New York, NY 10032
Tel: (+1) 212 342 0121
Fax: (+1) 212 342 0855
Email: [log in to unmask]
On 5/16/06, Denise Dörfel <[log in to unmask]> wrote:
> Dear SPM experts,
>
> I wanted to do a PPI for a contrast over 9 runs in an event related
> paradigm choosing a frontal ROI (FWE corrected).
> After specifying the PPI design and during assigning the data following
> error occured.
> It seems that there is no data in the ROI (design matrix looks like
> that), but that can't be.
> Any ideas?
>
> ___________________
> SPM2: spm_fmri_spm_ui (v2.52) 10:43:51 - 16/05/2006
> ========================================================================
> Mapping files : ...done
> Calculating globals : ...done
> Saving SPM configuration : ...SPM.mat saved
> Design reporting : Warning: Divide by zero.
> (Type "warning off MATLAB:divideByZero" to suppress this warning.)
> > In c:\programme\spm2\spm2\spm_DesMtx.m at line 646
> In c:\programme\spm2\spm2\spm_DesRep.m at line 626
> In c:\programme\spm2\spm2\spm_fmri_spm_ui.m at line 451
> Warning: Divide by zero.
> ??? Error using ==> svd
> NaN or Inf prevents convergence.
>
> Error in ==> c:\programme\spm2\spm2\spm_sp.m (sf_set)
> On line 1122 ==> [x.u, s, x.v] = svd(full(X),0);
>
> Error in ==> c:\programme\spm2\spm2\spm_sp.m
> On line 221 ==> varargout = {sf_set(X)};
>
> Error in ==> c:\programme\spm2\spm2\spm_SpUtil.m
> On line 275 ==> sX = spm_sp('Set',varargin{2});
>
> Error in ==> c:\programme\spm2\spm2\spm_DesRep.m
> On line 685 ==> else, est = spm_SpUtil('IsCon',varargin{2}.X); end
>
> Error in ==> c:\programme\spm2\spm2\spm_fmri_spm_ui.m
> On line 451 ==> spm_DesRep('DesMtx',SPM.xX,fname,SPM.xsDes)
>
> ??? Error while evaluating uicontrol Callback.
> ______________________
>
> Thanks,
> Denise.
>
> --
> Denise Doerfel
> Dipl.-Psych.
>
> Biopsychology
> Dresden University of Technology
> Zellescher Weg 17
> 01069 Dresden
> Germany
>
> Tel.: 0049 - 351 - 463 33708
> Fax.: 0049 - 351 - 463 37274
> mail: [log in to unmask]
> http://biopsychologie.tu-dresden.de>
--
Tobias Egner, PhD
fMRI Research Center
Columbia University
Neurological Institute, Box 108
710 West 168th Street
New York, NY 10032
Tel: (+1) 212 342 0121
Fax: (+1) 212 342 0855
Email: [log in to unmask]
===========End of original message text===========
--------------------------------
Darren R. Gitelman, M.D.
Cognitive Neurology and the Alzheimer's Disease Center
Northwestern Univ., 320 E. Superior St., Searle 11-470, Chicago, IL 60611
Voice: (312) 908-9023 Fax: (312) 908-8789
--------------------------------
|
