[Eeglablist] Some confusion about ERSPs/ITC precomputing
qin.pengmin at gmail.com
Tue Jul 10 19:02:30 PDT 2012
Thank you very much! Your talk is very useful.
I have a similar question. I make a study file, and then then use the
precompute channel measure to calculate the ERSP and ITC. In this study
file including 22 participants, each participants have two conditions. The
epoch for each condition is 1100 ms (-200 to 900). when I calculate the
ERSP and ITC, I use three kinds of parameters for cycles: [1 0], [1 1] and
[1 0.5]; frequency [3 45], baseline . When I check the results of these
three analysis, the frequency of the results are same: 3.00 3.93
5.15 6.76 8.86 11.61 15.23 19.97 26.18 34.32 45.00. Is
this right? How could I get the results with frequency in the following
way: 3 6 9 12 15 18 ......45? Thanks a lots!
2012/7/10 Makoto Miyakoshi <mmiyakoshi at ucsd.edu>
> Dear Mengyan,
> If you want to have 3 Hz to 30 Hz range after wavelet transform, the
> problem is that you have only -200 to 800 ms so that it is extremely
> difficult to have 3 Hz.
> If you use 3 cycles at 3 Hz, your window side is (1000ms/3Hz)*3cycles =
> 1000ms, meaning that 1000ms length data generates 1 datapoint after wavelet
> transform. So your epoch can allow to generate only 1 datapoint!
> You can either
> 1. prepare longer epoch (from -1000 ms to 2000 ms relative to stimulus
> onset) and use 3 Hz 3 cycle at the lowest: then you'll have -500 to 1500 ms
> after wavelet transform. You lose 500 ms from both sides of your epoch,
> because your window side is 1000 ms. It does not matter very much if your
> epoch overlaps each other if it is not too much.
> 2. Give up the lowest frequency range of 3 Hz and use 5 Hz instead with 2
> cycles (decreasing the cycle number reduces sensitivity though). Then
> you'll have 0 to 800 ms after wavelet transform. You lose 200 ms from both
> sides of your epoch since your window size is (1000ms/5Hz)*2cycles = 400ms.
> You can try No.2 immediately. Use the following options to see what
> 'cycles', [2 0.5], 'freqs', [5 30], 'nfreqs', 50
> 2012/7/9 诸梦妍 <bj12116 at gmail.com>
>> Dear eeglablist,
>> I'm just computing EEG data using EEGLAB. However, I met some problems
>> about pre-computing ERSPs/ITC (channel measures) and hope someone could
>> help me.
>> In my experiment, I had 43 subjects, 500Hz sampling rate. It was a 2 by 2
>> repeated measures design, and each condition had about 30 trials and the
>> same 43 channels for each subject after rejecting. I pre-processed the raw
>> data in Brain Vision Analyzer until I extracted epochs for each
>> condition which was -200 to 800ms. I exported the epoched data to EEGLAB,
>> re-saved it as .set and .fdt files for each condition each subject, and
>> created a STUDY using those data. Then I just used the GUI and tried to
>> pre-compute the data.
>> For ERP part, it worked well and got the same result with Analyzer. But
>> for ERSPs/ITS part, it didn't. Actually, I do not quite understand the
>> parameters of ERSPs/ITC processing, so I just used the default. However, I
>> found that it only computed 23.4Hz to 250Hz's data, which do not include my
>> interest Frequency(3 to 30Hz). I tried to add options in GUI like "
>> 'cycles',[3 0.5], 'nfreqs', 100, 'freqs',[3 30] ",but it turned out as a
>> wrong syntax.
>> When I used the default parameter, it showed in the MATLAB command window
>> like this below. I was just wondering what the red sentence means ( 1)
>> to 4) ) and hope that someone could tell me what should I do if I just
>> want it to compute 3Hz to 30Hz frequency.
>> Computing Event-Related Spectral Perturbation (ERSP) and
>> Inter-Trial Phase Coherence (ITC) images based on 31 trials
>> of 500 frames sampled at 500 Hz.
>> Each trial contains samples from -200 ms before to
>> 798 ms after the timelocking event.
>> Image frequency direction: normal
>> [?]1) Using 3 cycles at lowest frequency to 16 at highest.
>> [?]2) Generating 200 time points (-129.0 to 727.0 ms)
>> Finding closest points for time variableTime values for time/freq
>> decomposition is not perfectly uniformly distributed
>> [?]3) The window size used is 71 samples (142 ms) wide.
>> [?]4) Estimating 100 log-spaced frequencies from 23.4 Hz to 250.0 Hz.
>> Processing time point (of 200): 10 20 30 40 50 60 70 80 90 100 110 120
>> 130 140 150 160 170 180 190 200
>> Computing the mean baseline spectrum
>> Note: Add output variables to command line call in history to
>> retrieve results and use the tftopo function to replot them
>> Thanks in advance,
>> Zhu Mengyan
>> Mengyan Zhu
>> Psychology department, Peking University
>> Dormitory 2061, Building 48,No.5 Yiheyuan Road, Haidian District, Beijing
>> 100871, China
>> E-mail: bj12116 at gmail.com
>> 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
> Makoto Miyakoshi
> JSPS Postdoctral Fellow for Research Abroad
> Swartz Center for Computational Neuroscience
> Institute for Neural Computation, University of California San Diego
> 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
-------------- next part --------------
An HTML attachment was scrubbed...
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 385 bytes
Desc: not available
Url : http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20120710/7c70468d/attachment.gif
More information about the eeglablist