[Eeglablist] Electrodes Values

Tue May 28 09:10:34 PDT 2013

This discussion is missing the intention for summarizing the simultaneous values at all electrodes by a single number. Knowing the intended use would indicate whether the current mean is relevant or not. At the peak of a P300 or a CNV wave, do you want the mean to be ignored or not? As far as I remember, GFP discards the mean (i.e. applies the average reference before calculating power). If you consider that the mean is relevant, then you must provide a baseline for each channel. The system low-pass filter might be enough (i.e. you may consider that each channel fluctuates about its own zero baseline) but, depending on your intention, the mean value of each channel on some remote (but not too remote interval) might be better. In EPR studies, often the mean 1.0 or 0.5 s before stimulus onset is used. 

The choice of a quantifier must depend on what you want to infer from its use. If you intend a conclusion about the neural activation inside the brain, you would need to take into account how superficial the various sources are. This would get very complex. If you plan to compare children with adults or men with women under some conditions, then you may have to normalize your scales to equate every subject on some other condition, in order to account for attenuation of the scalp EEG by different skull thicknesses.

There should be no unique answer to the question initially asked.

André Achim,
Département de Psychologie
Université du Québec à Montréal

-----Message d'origine-----
De : eeglablist-bounces at sccn.ucsd.edu [mailto:eeglablist-bounces at sccn.ucsd.edu] De la part de Barry Oken
Envoyé : 22 mai 2013 00:59
À : Eriksen, Jeff :LPH EEG; Andreas Widmann; Arnaud Delorme
Cc : eeglablist at sccn.ucsd.edu
Objet : Re: [Eeglablist] Electrodes Values

For Andreas, the rationale for N-1 in much of statistics is that it gives a more unbiased estimate of the parameter in the whole population. If one really just wants to know the parameter in one particular sample, be it a single EEG or a single group of subjects without thinking about the population that it was sampled  from, then, as Jeff commented, N is correct.
Barry Oken

-----Original Message-----
From: eeglablist-bounces at sccn.ucsd.edu [mailto:eeglablist-bounces at sccn.ucsd.edu] On Behalf Of Eriksen, Jeff :LPH EEG
Sent: Tuesday, May 21, 2013 12:44 PM
To: Andreas Widmann; Arnaud Delorme
Cc: eeglablist at sccn.ucsd.edu
Subject: Re: [Eeglablist] Electrodes Values

All:

I believe we should use N, not N-1. This is a physical quantity, not a statistical estimate of some population parameter. If I want the RMS of a time series, I use every time sample and divide by the total number of time samples. If I want spatial global field power, I want to use all N spatial samples as well.

-Jeff Eriksen

-----Original Message-----
From: eeglablist-bounces at sccn.ucsd.edu [mailto:eeglablist-bounces at sccn.ucsd.edu] On Behalf Of Andreas Widmann
Sent: Tuesday, May 21, 2013 7:19 AM
To: Arnaud Delorme
Cc: eeglablist at sccn.ucsd.edu
Subject: Re: [Eeglablist] Electrodes Values

Dear Arno,

your notion of "small normalization difference" made me curious where this definition of GFP might originate. Where is it defined with "N - 1 normalization"? Also in ERPLAB GFP appears to be implemented according to this definition.

However, in the original paper by Lehmann and Skrandies (1980) it is either defined as "root of the mean of the squared potential differences between all possible electrode pairs" (formula 1A) gfp = sqrt(1 / (2 * 32) * sum(sum((repmat(tmp(:, 1), [1 32]) - repmat(tmp(:, 1), [1 32])') .^ 2))) (similar to implementation in eeg_gfp.m) or as root mean squared voltage of the average referenced data (formula 1B), with gfp_1A = gfp_1B * sqrt(nChans).

Also some other papers which explicitly define the term use the latter definition (e.g., Brandeis et al., 1992; Roth et al. 1995; Murray et al., 2008; Koenig et al., 2009), that is, normalize by N not N - 1. Shouldn't we possibly better use gfp = std(tmp, 1) % GFP normalized by N to compute GFP?

What is the rationale behind N - 1 normalization? Any ideas?
Thank you! Best,
Andreas

Am 18.05.2013 um 19:21 schrieb Arnaud Delorme <arno at ucsd.edu>:

> Dear Kleber,
> 
> one way to get a global measure proportional to microvolt is to use the Global Field Power (GFP).
> 
> http://sccn.ucsd.edu/pipermail/eeglablist/2008/002132.html
> 
> The root mean square (RMS) value is also sometimes used.
> The root mean square value on average reference data is similar to the GFP (with small normalization difference).
> 
> tmp = rand(32,10); % simulated data 32 channels 10 points
> std(tmp) % GFP
> sqrt(mean(tmp.^2)) % RMS
> sqrt(sum((tmp-repmat(mean(tmp), [size(tmp,1) 1])).^2)/(size(tmp,1)-1)) 
> % RMS on average reference with n-1 normalization (same as GFP)
> 
> I am not sure RMS on a common reference should be used. Any other comments on this appreciated.
> Thanks,
> 
> Arno
> 
> On 16 May 2013, at 15:15, Kleber de Aguiar wrote:
> 
>> Hello Everyone in this list,
>> 
>> I would like to know if anyone of you can answer me a doubt about de 
>> reading of the values of a group of electrodes in a EEG exam, as I"ll explain bellow:
>> 
>> 
>> 1- Spliting the EEG recording in slots of time, that I will call them 
>> as "samples";
>> 
>> 2- Assuming that "n" electrodes were used in this EEG Recording;
>> 
>> 3 - In every "sample" of the recording, there will have "n" 
>> electrodes values corresponding for each electrodes;
>> 
>> MY QUESTIONS:
>> 
>> * How can I sum this electrodes values of the sample in a single 
>> value that represents  the total of microvolts oh that "sample"?
>> 
>> * How can I describe the values of "y" samples in a single value that 
>> represents these "y" sample values?
>> 
>> 
>> Best regards,
>> 
>> Kleber, a fresh researcher in this field.
>> 
>> 
>> 
>> --
>> Programa de Engenharia de Sistemas e Computação PESC/COPPE/UFRJ 
>> Visite nossa página web (http://www.cos.ufrj.br)
>>       ------***------
>> Kleber de Aguiar
>> Tecnólogo em Sistemas de Computação
>> Mestrando em Engenharia de Sistemas e Computação - PESC/COPPE/UFRJ 
>> Tutor Presencial CEDERJ/UAB - Curso de Computação UFF
>> Twitter: @kleberIAguiar
>> 
>> 
>> _______________________________________________
>> 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
> 
> 
> _______________________________________________
> 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
> 

_______________________________________________
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

_______________________________________________
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

_______________________________________________
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