Dear Joe,
A quick question regarding time-frequency data. I see from the documentation that it is possible to create these data within the Toolkit, and that PCA can be performed on frequency data once generated. However I’ve not been able to ascertain whether time-frequency Eeglab files can be Read into the toolkit in the way that I have successfully done for time domain Eeglab files.
Since exporting Brain Vision time-frequency data to Eeglab is also presenting some challenges in itself, I thought I’d check whether I will be able to get such an Eeglab file into the Toolkit at the other end.
I appreciate that if a time-frequency Eeglab file cannot be read, I can instead create the data in the Toolkit from time domain files read in as Eeglab files (as per your documentation), I simply prefer to do analyses using single trials (which the initial time frequency transform entails) in Brain Vision Analyzer. I also appreciate that rather than trying to import a time-frequency file, I might alternatively import into the Toolkit each of the frequency bands (wavelets) in such a file simply as separate time domain files. These could be used as separate observations in a PCA, to make use of the frequency dimension, though the PCA would be “ignorant” of the existence of a frequency dimension and could not perform analysis, or plotting with respect to this.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Dear Joe,
A quick question regarding time-frequency data. I see from the documentation that it is possible to create these data within the Toolkit, and that PCA can be performed on frequency data once generated. However I’ve not been able to ascertain whether time-frequency Eeglab files can be Read into the toolkit in the way that I have successfully done for time domain Eeglab files.
Since exporting Brain Vision time-frequency data to Eeglab is also presenting some challenges in itself, I thought I’d check whether I will be able to get such an Eeglab file into the Toolkit at the other end.
I appreciate that if a time-frequency Eeglab file cannot be read, I can instead create the data in the Toolkit from time domain files read in as Eeglab files (as per your documentation), I simply prefer to do analyses using single trials (which the initial time frequency transform entails) in Brain Vision Analyzer. I also appreciate that rather than trying to import a time-frequency file, I might alternatively import into the Toolkit each of the frequency bands (wavelets) in such a file simply as separate time domain files. These could be used as separate observations in a PCA, to make use of the frequency dimension, though the PCA would be “ignorant” of the existence of a frequency dimension and could not perform analysis, or plotting with respect to this.
I never got around to implementing support for importing time-frequency EEGlab (or Brain Vision) but would be willing to give it a go if you could supply me with some example files. The EP Toolkit can perform analyses on single trials as well as averages if that is a concern.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Joe,
Thanks, and sorry for not replying sooner (new baby in house). For my purposes at present, i.e. complying with reviewers requests, I suspect that PCA of time frequency data will be overkill. Its possible I might have a go with a future project however. Brain Vision Analyzer does not support
the exporting of time frequency (wavelet) data to EEGLab, and having chatted with them about this, they did not come up with any solution other than a generic export of the time frequency data as a text file (with each frequency concatenated onto the others), and an accompanying header file. Obviously, prior to getting it into the EP Toolkit, this leaves open the legwork of reading that text file into EEGLab or similar, which I am no expert on. I’ve attached example files in case you feel this is worth pursuing in the development of the EP Toolkit, however do note that this is a rather specific fix, and not necessarily one that even I will be using immediately, so you may feel this belongs at the bottom of your to do list
The data started as 374 segments (trials, epochs), using sixty four electrodes in a standard montage (electrode labels are provided in the dat file, though this is optional at export), sampled from -100 to 700 ms at 500 Hz. These have been subjected to a time frequency transform to produce four wavelets at 1, 2, 3, and 4 Hz. The 374 segments were then averaged. This information is present in the header file
The structure of the data is simple, 64 rows for channels the first 400 columns are samples -100 to 700 for the 1 Hz frequency, the next 400 samples for 2 Hz, etc. Thus there are 1600 columns of data.
I took a closer look at EEGlab and it doesn't actually support representing time-frequency information in .set files anyway. Their approach is to just store the raw data and then generate the time-frequency data on the fly. It is possible to export the time-frequency info into a text file for further analysis using the command window. If you do that, it should be possible to then import the data into EP Toolkit using the Single File Mode, perhaps after some reformatting of the text files. It would be a hassle and I would instead recommend just performing the time-frequency transforms inside EP Toolkit.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Dear Joe,
A quick question regarding time-frequency data. I see from the documentation that it is possible to create these data within the Toolkit, and that PCA can be performed on frequency data once generated. However I’ve not been able to ascertain whether time-frequency Eeglab files can be Read into the toolkit in the way that I have successfully done for time domain Eeglab files.
Since exporting Brain Vision time-frequency data to Eeglab is also presenting some challenges in itself, I thought I’d check whether I will be able to get such an Eeglab file into the Toolkit at the other end.
I appreciate that if a time-frequency Eeglab file cannot be read, I can instead create the data in the Toolkit from time domain files read in as Eeglab files (as per your documentation), I simply prefer to do analyses using single trials (which the initial time frequency transform entails) in Brain Vision Analyzer. I also appreciate that rather than trying to import a time-frequency file, I might alternatively import into the Toolkit each of the frequency bands (wavelets) in such a file simply as separate time domain files. These could be used as separate observations in a PCA, to make use of the frequency dimension, though the PCA would be “ignorant” of the existence of a frequency dimension and could not perform analysis, or plotting with respect to this.
I never got around to implementing support for importing time-frequency EEGlab (or Brain Vision) but would be willing to give it a go if you could supply me with some example files. The EP Toolkit can perform analyses on single trials as well as averages if that is a concern.
Hi Joe,
Thanks, and sorry for not replying sooner (new baby in house). For my purposes at present, i.e. complying with reviewers requests, I suspect that PCA of time frequency data will be overkill. Its possible I might have a go with a future project however. Brain Vision Analyzer does not support
the exporting of time frequency (wavelet) data to EEGLab, and having chatted with them about this, they did not come up with any solution other than a generic export of the time frequency data as a text file (with each frequency concatenated onto the others), and an accompanying header file. Obviously, prior to getting it into the EP Toolkit, this leaves open the legwork of reading that text file into EEGLab or similar, which I am no expert on. I’ve attached example files in case you feel this is worth pursuing in the development of the EP Toolkit, however do note that this is a rather specific fix, and not necessarily one that even I will be using immediately, so you may feel this belongs at the bottom of your to do list
The data started as 374 segments (trials, epochs), using sixty four electrodes in a standard montage (electrode labels are provided in the dat file, though this is optional at export), sampled from -100 to 700 ms at 500 Hz. These have been subjected to a time frequency transform to produce four wavelets at 1, 2, 3, and 4 Hz. The 374 segments were then averaged. This information is present in the header file
The structure of the data is simple, 64 rows for channels the first 400 columns are samples -100 to 700 for the 1 Hz frequency, the next 400 samples for 2 Hz, etc. Thus there are 1600 columns of data.
I took a closer look at EEGlab and it doesn't actually support representing time-frequency information in .set files anyway. Their approach is to just store the raw data and then generate the time-frequency data on the fly. It is possible to export the time-frequency info into a text file for further analysis using the command window. If you do that, it should be possible to then import the data into EP Toolkit using the Single File Mode, perhaps after some reformatting of the text files. It would be a hassle and I would instead recommend just performing the time-frequency transforms inside EP Toolkit.