[Eeglablist] Low Pass Filtering and ERSP
mmiyakoshi at ucsd.edu
Thu Jan 12 20:23:09 PST 2017
EEGLAB's ERSP by default plots dB converted values, namely (post stimulus
power)/(baseline power). Therefore, in theory even if amplitude is
attenuated, their ratio of before vs. after the stimulus onset should
remain the same. However, I'm rather surprised that the ERSP plot seems
intact up to 100Hz!
As you can see, if that 20Hz noise has constant power before and after the
stimulus onset, they will cancel out after taking this ratio (which is why
you don't notice 50/60Hz line noise in ERSP.)
> In this case Notch filtering may not be necessary (as some people in my
lab have recommended).
Judging from the results, yes. Actually, the result rather seems to have
demonstrated that frequency filter does not work on ERSP results. So
applying notch filter would be meaningless.
> Because in either case I will be analyzing the lower frequencies..Have I
If you want to be rigorous, it could be still a good idea to limit your
freq band of interest below the noise frequency. However, it's not a bad
idea to include the broadband spectra if you don't see anything funny in
the final result (this means that you proved the stationarity of the 20Hz
noise, so they are invisible after baseline normalization by using dB.)
>From Lufthansa455 to Frankfurt,
On Fri, Dec 30, 2016 at 2:09 AM, Gozde BAYER <gbayer at hacettepe.edu.tr>
> Dear Makoto,
> Thank you very much for your reply.
> I have attached the ERSP analysis of channel C3 in jpg format. There are
> two conditions (hs,ss) from 7 subjects and the data were lowpass filtered
> at 20 Hz in the preprocessing step. As you may realize, there are still
> ersp data values just above 20 Hz (extending to about 30 Hz). I think it is
> due to the lowpass FIR filter that lowers the frequency amplitudes (above
> the filter value) slowly..But as you mentioned, the frequencies just below
> it (20 Hz) still seem intact. In this case Notch filtering may not be
> necessary (as some people in my lab have recommended). Because in either
> case I will be analyzing the lower frequencies..Have I understood
> Thank you very much again,
> Best wishes.
> Gözde BAYER
> On 30-12-2016 02:17, Makoto Miyakoshi wrote:
> Dear Gozde,
> If you can upload screenshot of the data in question, I can help you
> better. Let's identify what was wrong.
> Generally speaking, it would be still ok to apply 20-Hz low pass filter if
> you say only < 20 Hz of your data seems intact, just in the worse case.
> On Tue, Dec 13, 2016 at 11:21 PM, Gozde BAYER <gbayer at hacettepe.edu.tr>
>> Dear EEGLAB List,
>> During my raw EEG data recording from different subjects, I observed
>> (online from the FFT plot of related software) some kind of 'noise'
>> standing still at around 20-25 Hz and I could not get rid of that for some
>> reason. So I decided to let EEG stream under
>> this condition. Therefore I applied low pass filtering to the raw data at
>> 20 Hz, and analyzed offline the remaining frequencies with EEGLAB. I
>> created STUDY design for the purpose of the experiment. But when I plot
>> ERSP of a specific channel, I could still distinguish frequencies surviving
>> above 25 Hz. I checked from the workspace of Matlab as well. What could be
>> the reason of that? Since I am sometimes getting this 20 Hz peak during raw
>> EEG data recoding (by the way, I am trying to find out why this is the
>> case), it is important for me to really understand the phenomenon behind
>> that..Indeed, those frequencies just above 20 Hz are also of great
>> importance for my future analysis..If they are able to survive -somehow-
>> then would it be reliable to analyze beta frequency band under this
>> I appreciate any help,
>> Best wishes.
>> Gözde BAYER
>> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
>> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.uc
>> For digest mode, send an email with the subject "set digest mime" to
>> eeglablist-request at sccn.ucsd.edu
> Makoto Miyakoshi
> Swartz Center for Computational Neuroscience
> Institute for Neural Computation, University of California San Diego
Swartz Center for Computational Neuroscience
Institute for Neural Computation, University of California San Diego
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the eeglablist