Hi Bin,

Please comment out the lines 142-147 in spm_dcm_ind_data.m:

% check data are not oversampled (< 4ms)

%--------------------------------------------------------------------------

if (T2 - T1)/DT > 256

DT = fix((T2 - T1)/128);

end

DCM.options.D = DT;


I think what happened is that this downsampling reduced your time resolution and therefore the edge artefacts started to show up in your window of interest. This shouldn't happen if you remove these lines. The lines will be commented out in the next SPM release but you could do this in your own code for now.

Note that there is also another check on the sampling rate a few lines above in the code. This checks whether your sampling rate after downsampling is still higher than 250Hz, and if so, reduces it. If you don't want this to happen you could also comment out these lines and then the downsampling is exactly as you specify.

Let me know in case this did not solve the problem!

Best,
Bernadette


2014-09-29 22:19 GMT+01:00 Bin Wang <[log in to unmask]>:
Dear SPM&DCM Experts, 

I am encountering a problem when I do DCM estimation for EEG using SPM12. I downsampled the data to 200Hz and epoched the data (-1000ms, 2000ms) during the preprocessing using EEGLAB. After DCM inversion, I got some unreasonable time-frequecy outcomes using SPM12 (there are something look like edge artefacts at the beginning and end of time, see attachments). But I used the totally same data and parameters in SPM8,  I can get reasonable results of time-frequency plot. 
When I changed to use 2 in subsample option in SPM12, the results seem right, which should not be correct option because I have downsampled during preprocessing. I am wondering if anyone have the same problems? And, where did the problems happen? 

Any ideas or suggestions would be very appreciated. Thanks a lot in advance!

Bin


++++++++++++++++++++++++++++++
Bin Wang
Cognitive Neuroscience
Institute of Neuroscience and Medicine (INM-3)
Forchungszentrum Juelich, 
Leo-Brandt-Str. 5 
52425 Juelich
Email: [log in to unmask]
     &  [log in to unmask]  
Telefon: 02461-61-9542
++++++++++++++++++++++++++++++