Hi Christophe,
> I am trying to use SPM to analyse NIRS data.
> (the data are collected at 24 electrodes).
>
> For the moment, I am not interested in the spatial aspects:
> I just want a quick way to compute the fits between a model
> and the data (I am only interested in the voxel-based uncorrected p).
>
> I have managed to export the NIRS data in a series of img files that I
> have set up with the dimensions 2x5x5.
>
> When I try to estimate the model, the smoothness estimation crashes, which
> is not very surprising.
>
> I envisage two solutions:
> 1. to modify spm_spm.m to bypass the smoothness estimation, but this may
> open a can of worms (?)
> 2. to create larger img files so that the smoothness estimation has enough
> "space" to proceed.
>
> I would rather start with solution 2, but rather than proceeding by trial
> and errors, I am wondering if someone here could tell me what are the
> minimal requirements for the smoothness estimation to work (or propose
> another solution)
I tried solution 1 with 1-dimensional data a while ago. The only thing
one has to do with your data is to skip the smoothness estimation part
in spm_spm.m, but replace the final estimates (that are saved) after the
estimation loop by some dummy values. You have to modify FWHM, RESEL,
and R. Of course, all computations based on these values (corrected
p-values) will be wrong, but SPM shouldn't complain and happily compute
some corrected p-values.
Solution 2: I believe, it would be easier for you, if you arranged your
data in 2D-slices (similar to the 2D-representation of the 10-20 system
in EEG electrode placement), where the z-dimension is 1. For this case
(there were some emails on the helpline about this), SPM99 will properly
estimate smoothness estimates and corrected p-values for 2D-data. I
don't know, if there is a limitation on the minimal size of your images
with respect to the smoothness estimation. However, as you say, you can
always increase the size of your images (e.h. by interpolating them) and
get around such a limitation.
Stefan
--
Stefan Kiebel
Functional Imaging Laboratory
Wellcome Dept. of Cognitive Neurology
12 Queen Square
WC1N 3BG London, UK
Tel.: +44-(0)20-7833-7478
FAX : -7813-1420
email: [log in to unmask]
|