[Erppcatoolkit-support] Problem loading/appending data and question

[Andreas W. <wi...@un...>]

Dear Joe and list!

Thank you for providing your toolbox (and supporting the „Widmann variant“ of the EEGLAB file format; which I didn’t even know that it exists :)))! May I ask three short questions?

(1) I have a problem appending cells with the above mentioned file format in 2.47 and 2.48 but not in 2.45:
The error is:

Conversion to double from cell is not possible.
Error in ep_addData (line 287)
                EPdataOut.trialSpecs(end+1:end+numAdded,:)=cell(numAdded,size(EPdataIn.trialSpecs,2));
Error in ep_editData (line 2666)
            EPoverview.workData=ep_addData(EPoverview.workData,newData,'cells'); 
Error while evaluating uicontrol Callback

If I add a check in ep_addData whether EPdataIn.trialSpecs is empty (as it was in previous versions) it works:
if ~isempty(EPdataIn.trialSpecs)
                EPdataOut.trialSpecs(end+1:end+numAdded,:)=cell(numAdded,size(EPdataIn.trialSpecs,2));
end

Would you expect any unexpected side effects from this workaround? Why did you remove the check? (Btw: Is there a way to use the single cell file mode for reading this file format? Couldn’t get that to work. Most probably due to my ignorance.)

(2) Only short suggestion: In ep_addEloc, line 134 ...
        if ~isempty(eloc(i).labels) && ~isempty(eloc(i).theta) && isempty(eloc(i).type)
… it could be helpful to check whether theta is numeric before wiping. EEGLAB pop_chanedit always ignores type when importing ced-files even if it is contained in the file. This practically disables importing electrode locations if one does not know that the type field has to be filled manually.

(3) My main question: I would like to compare the oddball P3a component latency with the latency of some other external non-ERP psychophysical measurement. Usually, we would do this using jackknifing with some relative peak amplitude criterion. However, I would prefer doing this with a „true" PCA-P3a component and not the ERP-P3a. I cannot separate N1/P2/P3a components with spatial PCA as they are too similar. Now, I hope my question is not too stupid: Do you think it would be a valid approach to do so some kind of jackknifing temporal PCA? That is, performing several temporal PCAs omitting one subject and estimating jackknifing latency from each of the resulting PCA-P3a component time courses?

Thank you! Best regards,
Andreas