Agatha
Here is some history for the problem:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0708&L=SPM&P=R11909
The bottom line is that you can try changing -32 to -16 in the last
line of spm_PEB
-----
Darren Gitelman
On Wed, Dec 31, 2008 at 2:50 PM, Agatha Lenartowicz <[log in to unmask]> wrote:
> Hi all ~
> I am trying to figure out if an error message (matlab default warning)
> occurring during a matrix inversion in the spm_PEB.m script is benign or of
> concern (the message is below).
> Note that the spm_PEB.m script is being called from within spm_peb_ppi.m -
> so the design struct and data (ROI timeseries) are atypical (the script is
> being used here to obtain estimates to be used for deconvolution of the BOLD
> signal timeseries).
> Though I can navigate within the guts of spm_PEB, I am not quite fluent in
> parametric empirical Bayesian estimators. And so I would be grateful if
> anyone could either:
> - point me to the variables/matrices produced by the spm_PEB script (matrix
> of interest?) that would be helpful in diagnosing the issue or..
> - take a look at the input structure P which I am feeding into the spm_PEB.m
> script (C = spm_PEB(Y,P);)
> Many thanks and Happy 2009!
> Agatha ~
>
> ERROR:
> ----------
> Warning: Matrix is close to singular or badly scaled.
> Results may be inaccurate. RCOND = 1.266417e-20.
> In spm_PEB>spm_inv at 371
> In spm_PEB at 244
> ----------
>
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Post Doctoral Scholar
> Laboratory of R.A. Poldrack
> www.poldracklab.org
> UCLA Department of Psychology
> 1285 Franz Hall, Box 951563
> Los Angeles, CA, 90095-1563
> http://alenarto.bol.ucla.edu/
>
>
>
>
|