Dear Andreas,
I've re-run my analysis of the face data but
have been unable to replicate your bug.
spm_reml ran fine.
To help us solve this problem could you send your SPM.mat
file* to Karl, who will then have a look at this.
Best wishes,
Will.
*Or better still, if you type dbstop if error and then
save the variables that are passed to spm_reml.m in a file
'for_karl.mat' and then send just these to him - that would
be optimal !
Will Penny wrote:
> Dear Karl,
>
> There seems to be a problem with ReML.
>
> This time, from an fMRI analysis of Rik's face repetition data.
>
> This estimated fine when I wrote the manual. And Rik
> repeated the analysis a week or two ago. So it must be a
> recent development.
>
> Perhaps you could help ?
>
> Thanks,
>
> Will.
>
> -------- Original Message --------
> Subject: Re: [SPM] Estimation problem in Face data set
> Date: Wed, 1 Mar 2006 16:15:28 -0500
> From: Walsh, Nicholas <[log in to unmask]>
> Reply-To: Walsh, Nicholas <[log in to unmask]>
> To: [log in to unmask]
>
> Yes - I'm still stuck at this stage and it is unclear where to go next.
>
> Any help and advice from the SPM manual authors would be great as to
> whether it is a bug.
>
> Nick
>
> -----Original Message-----
> From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]
> On Behalf Of Andreas Bruns
> Sent: Wednesday, March 01, 2006 10:29 AM
> To: [log in to unmask]
> Subject: [SPM] Estimation problem in Face data set
>
> Dear All,
>
> I encountered the same problem as Juan and Nick (both 24 Feb, "Help with
>
> estimation problems", "Error when estimating model"), and moreover, it
> shows up while processing the "Face" example data set (SPM5 manual,
> chapter
> 29). After successful spatial processing and model specification, the
> error
> occurs during model estimation (section 29.2.1):
>
> ********************************
>
> Initialising parameters :
> ...initialised
>
> Plane 1/46 , block 1/9 : ...read & mask data
> Plane 1/46 , block 2/9 : ...read & mask data
> Plane 1/46 , block 3/9 : ...read & mask data
> .
> .
> .
> Plane 46/46 , block 7/9 : ...read & mask data
> Plane 46/46 , block 8/9 : ...read & mask data
> Plane 46/46 , block 9/9 : ...done
> Temporal non-sphericity (over voxels) : ...REML estimation
> ReML Block - 1
> ReML Iteration : 1 ...NaN
> Warning: Matrix is close to singular or badly scaled.
> Results may be inaccurate. RCOND = -1.000000e+00.
>
>> In spm_reml at 90
>
> In spm_spm at 855
> In spm_config_fmri_est>run_est at 382
> In spm_jobman>run_struct1 at 1384
> In spm_jobman>run_struct1 at 1392
> In spm_jobman>run_struct1 at 1392
> In spm_jobman>run_struct at 1351
> ??? Error using ==> svd
> NaN or Inf prevents convergence.
>
> ??? Error while evaluating uicontrol Callback.
>
> ********************************
>
> The problem obviously is that in 'spm_reml', lines 76ff., the matrix 'C'
>
> looks "unsuspicious" in the 1st iteration (k=1), but contains merely
> NaN's
> around the diagonal in the 2nd iteration (k=2):
>
>>> num2str(full(C(1:10,1:10)),'% 8.3f')
>
> ans =
> NaN NaN NaN NaN NaN NaN 0.000 0.000 0.000
> 0.000
> NaN NaN NaN NaN NaN NaN NaN 0.000 0.000
> 0.000
> NaN NaN NaN NaN NaN NaN NaN NaN 0.000
> 0.000
> NaN NaN NaN NaN NaN NaN NaN NaN NaN
> 0.000
> NaN NaN NaN NaN NaN NaN NaN NaN NaN
> NaN
> NaN NaN NaN NaN NaN NaN NaN NaN NaN
> NaN
> 0.000 NaN NaN NaN NaN NaN NaN NaN NaN
> NaN
> 0.000 0.000 NaN NaN NaN NaN NaN NaN NaN
> NaN
> 0.000 0.000 0.000 NaN NaN NaN NaN NaN NaN
> NaN
> 0.000 0.000 0.000 0.000 NaN NaN NaN NaN NaN
> NaN
>
> However, before going deeper into the matter, I would like to know
> whether
> this might be an SPM5 bug, since the problem already occurs in one of
> the
> example data sets. I am working with MATLAB R14SP3 under Linux and am
> using
> SPM5 revision no. 456.
>
> Any help is greatly appreciated. Best regards,
>
> Andreas
>
> ------------------
> Dr. Andreas Bruns
> Biozentrum, University of Basel
> Klingelbergstrasse 50/70
> 4056 Basel
> Switzerland
> ------------------
>
>
>
>
--
William D. Penny
Wellcome Department of Imaging Neuroscience
University College London
12 Queen Square
London WC1N 3BG
Tel: 020 7833 7475
FAX: 020 7813 1420
Email: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
|