[Eeglablist] unsure what i'm doing wrong...

[Andrew Hill]

Hi folks,

If someone could point me in the right direction.. 
I keep running into blockers trying to get through the ICA workflow.  
The latest one is baffling me.

At the moment here is my workflow:

First script:
1. import .bdf files to .sets, filter pass .16 to 50 hz
2. detrend the bdf recordings
3. remove any bad channels with pop_rejchan
4. pop_runica with default options
5. extract event-specific epochs with pop_epoch, saving a file for each event type
6. pop_autorej to reject any bad epochs automagically, pop_editset for the group and session info

Second script:
Creates a study with the relevant events to compare (depending on what I'm looking at, I've got several different studies composed of different event-based epochs).

The study creates successfully, with:
Nb of subjects: 40
Nb of conditions: 2 per subject
Nb of sessions: 2 per subject.
Nb of groups : 4 (some missing) I assume this is because my groups have unequal numbers of subjects.
Epoch consistency: yes
Channels per frame: 58,59,60,61,62,63,64,65,66  I assume because pop_rejchan drops different channels in each recording.
Clusters: 1
Status is "Ready to precluster"

One concern is when creating the study, after EEGLAB reads in each subject I get warnings:

eeg_checkset: recomputing the ICA activation matrix ...
Warning: ***** change STUDY design as it combines datasets with different ICA decompositions
Warning: ***** change STUDY design as it combines datasets with different ICA decompositions
Rebuilding trial information structure for STUDY

Now, after reloading a specific study I'm trying to precompute channel measures, interpolating missing channels at this time.  
I'm using the following for time/freq parameters:  'cycles', [3 0.5], 'nfreqs', 100, 'freqs',[0 50]

The process gets past the first subject and then fails with the message:

EEGLAB error in function std_spec() at line 295:
Dot name reference on non-scalar structure.
(If you think this is a bug ... etc).


As far as I understand, after precomputing/interpolating channel measures, I should precompute component measures, then build preclustering array, cluster components, and then edit/plot components... but I cannot get beyond the above error message.

So.... if I wasn't bald already I'd be pulling my hair out.
Any suggestions?

Thanks,
Andrew