Dear SPMers,
I am beginning to use SPM and DCM and I have a few remedial questions
about DCM.
We have analyzed some PPI data that suggests a relationship between regions
and we would like to use DCM to establish a model for connectivity. I've used
the example photic/motion/attention model that you've published on your
website, I've looked at the old ppts that you've posted from your yearly
courses on SPM, and I think I have a much better understanding of the theory
behind DCM. But I'm having some problems applying it:
I'm having trouble designing a DCM model. We have a fingertap experiment in
which we ask subjects to press a button in response to a stimulus. The
stimulus conditions vary - we have a regularly timed stimulus (beat), a
randomly timed stimulus (random), and a rest period in which there is no
stimulus (rest).
1)Does this seem like an appropriate way to specify the regressors for a DCM
analysis?
I then find the VOIs. I specify a DCM using my VOIs and ask SPM to estimate
a model using the connections I specify. MATLAB returns several errors after
calculating the first PEB and GNS iterations:
Warning: Returning NaN for out of range arguments
In spm_Ncdf at 76
In spm_dcm_estimate at 72
In spm_dcm_ui at 312
This first GNS iteration value is functionally zero (e-22). As a consequence,
when I review the DCM my modelled system output is flat.
2)What does this mean, why do I get such low values and why does the
program return an NaN? Is it a function of the data? Does it mean that I've
designed the model inappropriately? Or I'm not properly defining the DCM?
Any help is greatly appreciated. Thank you in advance.
|