Hi David,

I meant changing the serial correlation option in fMRI model
specification from AR(1) to None. If you are not making any inference,
it shouldn't really matter and model estimation might subsequently be
twice as fast.

Best regards,
Guillaume.


On 08/12/16 21:49, David Hofmann wrote:
> Hi Guillaume,
>
> thank you very much! I made all the changes and it is much fast now. One
> thing I didn't understand, though:
>
> What you mean by change "non-sphericity", i.e. what exactly should I
> change and where?
>
> greetings
>
> David
>
>
> 2016-12-07 12:11 GMT+01:00 Guillaume Flandin <[log in to unmask]
> <mailto:[log in to unmask]>>:

>
>     Dear David,
>
>     I assume that you already changed the defaults stats.resmem and
>     non-sphericity. Then, if you want to select the output files to write to
>     disk, you will have to modify the spm_spm.m function. For example, look
>     at the loop:
>
>         for j=1:nBeta
>             c(cmask) = beta(j,:);
>             Vbeta(j) = spm_data_write(Vbeta(j), c, chunk);
>         end
>
>     if you only want to write beta images 5 to 10, then change the for loop
>     to this:
>
>         for j=5:10
>
>     Headers for all images will still be written but they are small (you can
>     also choose which ones to write by editing line:
>
>         Vbeta = spm_data_hdr_write(Vbeta);
>
>     Best regards,
>     Guillaume.
>
>
>     On 05/12/16 16:32, David Hofmann wrote:
>     > Hi all,
>     >
>     > I'm estimating a trial-by-trial GLM for a few hundred subject and
>     > several tasks. SPM writes a lot of nifti files to the hard drive which
>     > takes up a lot of space and also prolongs the estimation due to the
>     > writing process. Since I am only interested in a few nifti files (e.g.
>     > don't need the other files for the motion regressors), I don't
>     need SPM
>     > to create all the other Nifti files.
>     >
>     > So I was wondering:
>     >
>     > 1. Is it possible to restrict SPM to only creating some nifti
>     files and
>     > don't write the others?
>     >
>     > 2. Is it possible to store the data as a mat-file instead of nifti
>     files?
>     >
>     > Any help appreciated!
>     >
>     > greetings
>     >
>     > David
>
>     --
>     Guillaume Flandin, PhD
>     Wellcome Trust Centre for Neuroimaging
>     University College London
>     12 Queen Square
>     London WC1N 3BG
>
>

--
Guillaume Flandin, PhD
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG