Modification needed for function epoch4bin (at least in erplab2023)

[gc0000]

Consider my data epoched with ERPLAB with a part of EEG.epoch variable shown in the picture. Obviously, there are a number of epochs belonging to bin 7, but epoch4bin(EEG,7) will return an empty [].

It is observed that when there is(are) one or more epoch(s) with 2 or more events defined in BDF, both EEG.epoch.eventbini and EEG.epoch.eventlatency variables will be the cell type. However, when there isn't, EEG.epoch.eventbini will be the double type, as shown in the picture (EEG.epoch.eventlatency will be 0s but not the cell type as well, but this doesn't matter in this issue). This very condition is not considered in the function epoch4bin.

Consider how epoch4bin will process the 1st epoch in the picture:

Let me first give value binArray=7; i=1;

then test the script as in epoch4bin
` bini = EEG.epoch(i).eventbini; % bin indices at this epoch
laten = EEG.epoch(i).eventlatency; % latencies at this epoch

  if iscell(laten)
        laten = cell2mat(laten);
  end
  
  indxtimelock = find(laten == 0,1,'first'); % catch zero-time locked code position,
  bin = bini(indxtimelock); % bin of the home item event code (time-locked event)
  
  if iscell(bin)
        bin = cell2mat(bin); %bug. It said "bini"
  end     
  if nnz(ismember_bc2(bin, binArray)) > 0
        indx(k) = i;
        k=k+1;
  end`

The result will be that nnz(ismember_bc2(bin, binArray))==0, because the sentence bin = bini(indxtimelock); has acted as if EEG.epoch.eventbini is a cell. Unfortunately, here bini is not a cell but a double [2, 7, 11]. Therefore bin will be 2 but not [2, 7, 11], then nnz(ismember_bc2(bin, binArray))==0. The epoch4bin function then will wrongly decide that my 1st epoch does not belong to bin 7.

Therefore, I suggest a modification of epoch4bin as shown below:

change bin = bini(indxtimelock);

to

if iscell(bini) bin = bini(indxtimelock); elseif isnumeric(bini) bin=bini; end

The EEG EVENTLIST example and my BDF txt file are attached.

13_Eventlist_Bins_TF.txt

BDFpare.txt

[Kwinsler]

Hello, thank you for the detailed report and proposed modification.

I am looking into this issue, but am having trouble replicating the case when EEG.epoch.evenbini contains doubles rather than cells. From what I understand, you are describing the case when there is a single event in an epoch, and it is in multiple bins. When I bin/epoch data like this, it results in a 1x1 cell, not a double. In all cases I've tested - epochs with one/multiple event(s) which may be in one/multiple bin(s) - EEG.epoch.evenbini contains 1xN cells, not doubles.

I might implement your suggested change regardless, but I would like to understand how this issue has come about, since it may affect other functions as well.

Could you share which version of MATLAB, EEGLAB, and ERPLAB you are using?

Could you share your procedure for epoching your data?

[gc0000]

Dear Kwinsler,

Sorry for the delay.

Please see https://github.com/gc0000/files-for-issues-of-erplab

The bug can be replicated with the 13_ICA_Cleaned_TF.mat file and the processes below (after the file is loaded to Matlab and eeglab is called):

BoundaryNum=-99; EpochLat=[-550,1250]; bl='none'; refpath='/a/path/containing/BDFpare.txt'; savepath='/an/existing/path/'; EEG = pop_creabasiceventlist( EEG , 'AlphanumericCleaning', 'on', 'BoundaryNumeric', { BoundaryNum }, 'BoundaryString', { 'boundary' }, 'Eventlist', [savepath '13_Eventlist_TF.txt']); EEG = pop_binlister( EEG , 'BDF', [refpath filesep 'BDFpare.txt'], 'ExportEL', [savepath '13_Eventlist_Bins_TF.txt'], 'IndexEL', 1, 'SendEL2', 'EEG&Text', 'UpdateEEG', 'on', 'Voutput', 'EEG' ); EEG = pop_epochbin( EEG , EpochLat, bl);

The Matlab version is R2023a, the eeglab version is v2022.0 and erplab version v9.00.

Note that some other commands I used (those for artifacts rejection) are not included for simplicity, and it does not matter in this case.

The 13_Epoch_Bin_Elist_TF.mat and the 13_Eventlist_Bins.txt file is the output of my own scripts.

My personal guess is that the bug occurs when and only when all epochs do contain and contain only 1 mark. You may create several pseudo-EEG-datas or read the pop_epochbin function to test my hypothesis. Another modification suggested, besides the above one, is to check the mechanism of outputting the double type but not cell type variables in EEG.epoch.eventbini when each epoch has only one mark. Such mechanisms may exist in pop_epochbin, which you may change instead (of epoch4bin)