[Eeglablist] study file issue when analyzing two IVs

[Delorme, Arnaud]

Dear Fran,

The error arises because you have participants for which you have 2 (or more) conditions and 2 (or more) different ICA decompositions. This is not allowed. When you are comparing 2 or more conditions within participants, the ICA decomposition must be the same for all conditions - so ICA must be run on a dataset which contains all the conditions and then you can separate the different conditions.  Even if you cluster ICA components, difference between conditions within participant might arise because of difference in the ICA decompositions - which is not what you want. If you were to see a difference it could be due to:
1- difference between ICA decompositions due to:
	- random initialization of the ICA algorithm
	- different level of noise in the data for each condition
	- different artifacts in the data for each condition
	- poor clustering to find corresponding components across conditions
2- genuine difference between brain source activity in the 2 conditions

By enforcing the same ICA decomposition across conditions, you remove all the potential differences due to 1 above so the data is easier to interpret. In other words, because you use the same spatial filter for both conditions, you can be sure that any difference you observe between conditions is not due to difference in the ICA spatial filter.

Note that when plotting data channels, it is allowed to remove artifacts even if the ICA decomposition is different across conditions within the same subject.

Cheers,

Arno

Ps: with respect to Makoto’s message, I do not think this is the problem, but to figure out where the data might be missing, below is the call to plot the data (for example).

STUDY = std_erpdata(STUDY,ALLEEG,'channels',{‘Cz'}, 'design', 1);

You can add an output to get the plotted data

[STUDY, erpdata] = std_erpdata(STUDY,ALLEEG,'channels',{‘Cz'}, 'design', 1);

If you cannot plot, you can simply read the data and assess where there might be a problem

[STUDY, erpdata] = std_readerp(STUDY,ALLEEG,'channels',{‘Cz'}, 'design', 1);

erpdata =

  2×1 cell array

    {750×3 single}
    {750×3 single}

figure; plot(erpdata{1}(:,1));


> On Jul 18, 2020, at 1:21 PM, Makoto Miyakoshi <mmiyakoshi at ucsd.edu> wrote:
> 
> Dear Fran,
> 
> My experience with the latest version of EEGLAB is limited, but I hope my
> general guesses helps.
> 
> 1. Most likely, you have empty data in one of the conditions in your
> balanced design. Check the data for an empty entry.
> 2. Condition name issues--do not use anything other than alpha-numerical
> characters (and for safety, do not use numbers to start with)
> 3. Similar to above, confusion between numbers and characters--sometimes it
> causes problems.
> 
> Makoto
> 
> 
> On Tue, Jul 14, 2020 at 1:52 PM Fran Copelli <fcopelli at ryerson.ca> wrote:
> 
>> Hi,
>> 
>> I have 44 participants (22 GroupA, 22 GroupB). All participants were
>> exposed to the same experiment (9 condition types) and preprocessed in the
>> same way.
>> 
>> I am trying to visualize ERSPs, but when I try to compare the two groups, I
>> receive the error *"cannot handle conditions with different number of
>> components, (error occurred in function std_readdata() at line 290)"*
>> 
>> I am sure that all participants have trials in the 9 conditions, so it is
>> not an issue of missing data. The study generates ERSPs when looking at
>> type alone (toggled in the '*edit study design*' menu),  but when
>> generating for  groups, I get the error.
>> 
>> Please advise as this is basic functionality of the eeglab study. I have a
>> colleague who has groupxcondition comparison, and is analyzing in electrode
>> space (as opposed to source analysis, like myself), and she was able to
>> generate her ERSPs just fine.
>> 
>> Thanks
>> 
>> *Fran*
>> 
>> SMART Lab, Psychology Department
>> _______________________________________________
>> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
>> To unsubscribe, send an empty email to
>> eeglablist-unsubscribe at sccn.ucsd.edu
>> For digest mode, send an email with the subject "set digest mime" to
>> eeglablist-request at sccn.ucsd.edu
>> 
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to eeglablist-request at sccn.ucsd.edu