Dear Professor Dien
I am a newer with ERP and PCA analysis, while I was doing data analysis with my experiment, I found a paper using this toolbox to do the PCA analysis,which enlightened my data results.
In this paper, the author described her process of PCA as follows:
Following Dien et al. (2007)’s recommendations for principal components analysis of ERP data, we conducted a temporal PCA followed by a spatial independent components analysis (ICA) of each temporal component. ...Mean voltage data was extracted from two spatial regions of interest (ROI) defined by the selected temporospatial factor from the principal components analysis (TF01SF1-Island PCA; TF01SF1-Non-island PCA) over a time window defined with a 0.6 beta weight criterion... This gave a frontal ROI with a time
window of 588–792 ms for the Island PCA and a posterior ROI with a time window of 612–796 ms for the Non-island PCA.
I have dowload the software from this website and read the tutorial varefully, following the tutorial, I have learned how to get the ROI through "windowing PCA data" with a 0.6 beta weight criterion, but still I have got a clue about how to get the time window through the software, so could you be patient to tell me how to identify time window with EP Toolkit or which part it is in the tutorial. Attached is the paper's supplementary about the data processing procedure.
Your reply will be very important for me and I really think this is a wonderful software.
Thanks for your kind words! I'm afraid I don't really understand your question. You set a window in the Windowing function. You enter in the start and end ms of the window you want.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Dear Professor Dien
I am sorry that I didn't clarify myself clearly.
In the paper, as a quoted “Mean voltage data was extracted from two spatial regions of interest (ROI) defined by the selected temporospatial factor from the principal components analysis (TF01SF1-Island PCA; TF01SF1-Non-island PCA) over a time window defined with a 0.6 beta weight criterion... This gave a frontal ROI with a time window of 588–792 ms for the Island PCA and a posterior ROI with a time window of 612–796 ms for the Non-island PCA.”
I understood it as with a 0.6 beta weight, the EP toolkit had selected the time window as 588-792 ms for TF01SF1-Island PCA and 612–796 ms for the Non-island PCA. that's to say, I thought the EP toolkit not only will identify ROI, but also tell us TF01SF1-Island PCA is most significant within the window 588-792 and TF01SF1-Non-island PCA is significant during 612–796 ms.
or I just misunderstood this sentence, only the ROI is identified by EP toolkit, the time window is selected by the author through other methods?
thank you again for your patience.
Yours Ji Xiaoli
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
No problem. Once the temporal PCA has been performed, it no longer matters what time points are chosen for the window. All time points reflect the same factor scores. You'll get the exact same p-values no matter what time points you choose. Please see my 2004 and 2012 papers for more explanation of the math behind this process.
Joe
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Dear Professor Dien
I read the tutorial again, on the page 130, the introduction about " Windowing Subject Average Data", the introduction said "The time window also needs to be specified. Let's try a 176-224 ms window (although this tutorial dataset is too small to look for significant effects) by entering in samples 70 to 81". my question is how this time window "176-224 ms" was selected for tf04sp01?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
In the case of the tutorial, this was handled by choosing the "autoPCA" option. It chooses the peak time point as the time window, which in this example was "192-196 ms." The "176-224 ms" you are quoting was for the windowing of the averaged data, not the PCA data. That would be chosen in the normal manner. In this case, the rationale was 25 ms on either side of 200 ms, for a 50ms long window, which is reasonable for an N2. If you're wondering why the numbers aren't multiples of five, it is because the samples are four ms long (250 Hz sampling rate), so all the window times must necessarily be multiples of four, so it was necessary to nudge the edges of the window to coincide with the nearest sample boundaries. When writing manuscripts, authors often ignore the reality that their window boundaries are constrained by the temporal resolution of their sample durations, giving a misleading impression of greater temporal precision than is actually present.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
thank you very much for your reply, I have a further question, is there any method to get the exact value of beta weight for every tempo factors over the time window?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Dear Professor Dien
I am a newer with ERP and PCA analysis, while I was doing data analysis with my experiment, I found a paper using this toolbox to do the PCA analysis,which enlightened my data results.
In this paper, the author described her process of PCA as follows:
Following Dien et al. (2007)’s recommendations for principal components analysis of ERP data, we conducted a temporal PCA followed by a spatial independent components analysis (ICA) of each temporal component. ...Mean voltage data was extracted from two spatial regions of interest (ROI) defined by the selected temporospatial factor from the principal components analysis (TF01SF1-Island PCA; TF01SF1-Non-island PCA) over a time window defined with a 0.6 beta weight criterion... This gave a frontal ROI with a time
window of 588–792 ms for the Island PCA and a posterior ROI with a time window of 612–796 ms for the Non-island PCA.
I have dowload the software from this website and read the tutorial varefully, following the tutorial, I have learned how to get the ROI through "windowing PCA data" with a 0.6 beta weight criterion, but still I have got a clue about how to get the time window through the software, so could you be patient to tell me how to identify time window with EP Toolkit or which part it is in the tutorial. Attached is the paper's supplementary about the data processing procedure.
Your reply will be very important for me and I really think this is a wonderful software.
Thanks for your kind words! I'm afraid I don't really understand your question. You set a window in the Windowing function. You enter in the start and end ms of the window you want.
Dear Professor Dien
I am sorry that I didn't clarify myself clearly.
In the paper, as a quoted “Mean voltage data was extracted from two spatial regions of interest (ROI) defined by the selected temporospatial factor from the principal components analysis (TF01SF1-Island PCA; TF01SF1-Non-island PCA) over a time window defined with a 0.6 beta weight criterion... This gave a frontal ROI with a time window of 588–792 ms for the Island PCA and a posterior ROI with a time window of 612–796 ms for the Non-island PCA.”
I understood it as with a 0.6 beta weight, the EP toolkit had selected the time window as 588-792 ms for TF01SF1-Island PCA and 612–796 ms for the Non-island PCA. that's to say, I thought the EP toolkit not only will identify ROI, but also tell us TF01SF1-Island PCA is most significant within the window 588-792 and TF01SF1-Non-island PCA is significant during 612–796 ms.
or I just misunderstood this sentence, only the ROI is identified by EP toolkit, the time window is selected by the author through other methods?
thank you again for your patience.
No problem. Once the temporal PCA has been performed, it no longer matters what time points are chosen for the window. All time points reflect the same factor scores. You'll get the exact same p-values no matter what time points you choose. Please see my 2004 and 2012 papers for more explanation of the math behind this process.
Joe
Dear Professor Dien
I read the tutorial again, on the page 130, the introduction about " Windowing Subject Average Data", the introduction said "The time window also needs to be specified. Let's try a 176-224 ms window (although this tutorial dataset is too small to look for significant effects) by entering in samples 70 to 81". my question is how this time window "176-224 ms" was selected for tf04sp01?
In the case of the tutorial, this was handled by choosing the "autoPCA" option. It chooses the peak time point as the time window, which in this example was "192-196 ms." The "176-224 ms" you are quoting was for the windowing of the averaged data, not the PCA data. That would be chosen in the normal manner. In this case, the rationale was 25 ms on either side of 200 ms, for a 50ms long window, which is reasonable for an N2. If you're wondering why the numbers aren't multiples of five, it is because the samples are four ms long (250 Hz sampling rate), so all the window times must necessarily be multiples of four, so it was necessary to nudge the edges of the window to coincide with the nearest sample boundaries. When writing manuscripts, authors often ignore the reality that their window boundaries are constrained by the temporal resolution of their sample durations, giving a misleading impression of greater temporal precision than is actually present.
thank you very much for your reply, I have a further question, is there any method to get the exact value of beta weight for every tempo factors over the time window?
If you mean the values of the factor pattern and factor structure matrices, you can obtain them using the Edit function.