[Eeglablist] EEGLAB Orientation

Mon Dec 16 12:46:07 PST 2024

Hi Makoto and Joe,

Sorry for slow response, I am just seeing these now. 

My recommendation of the best approach is actually to use Cyril's LIMO-EEG since it incorporates not only weighted least squares (WLS) optimization that takes care of outlier trials in a probably better way that is data-driven (as opposed to having to decide between 5, 10, 20% trim arbitrarily), and then applies Yuen t-tests (t-tests analysis) which implement 20% trim in their designs (here that would be a group level if I recall correctly). Plus, hypothesis testing is done using nonparametric statistics (1,000 iterations bootstrap + spatiotemporal clustering corrections for type 1 error). The whole thing seems like the most robust approach to me to deal with within and between subject variance/noise while not having to do excess data cleaning with arbitrary methods. 

Cedric Cannard

On Sunday, December 8th, 2024 at 7:44 PM, Joseph Dien via eeglablist <eeglablist at sccn.ucsd.edu> wrote:

> Agreed! It definitely depends on the data and the application. I tend
> towards a more conservative 5% trimming for robust ANOVAs, in part
> because folks still seem skittish about them. I think it is a bit odd
> that there is so much concern about a "replication crisis," with so
> little attention paid to the severe and known problems with the
> conventional ANOVA tests themselves. My EP Toolkit makes robust ANOVAs
> easy to use, but I see very little awareness of them. The field
> continues to be mired in 1960's statistics. Sigh...
> 
> Joe
> 
> On 12/6/24 10:11, Makoto Miyakoshi via eeglablist wrote:
> 
> > Hi Joe,
> > 
> > Thank you for your valuable input. I definitely agree with you, based on my
> > non-systematic observations too.
> > 
> > It is good to learn that the developers of robust ANOVAs recommended 20%
> > trimming and your recommendation of 5% trimming.
> > 
> > I think there is a psychological factor here; how much robustness one
> > requires reflects how deeply traumatized one is from past experiences of
> > EEG processing. I tend to be on the side of safety, and I have published
> > EEG datasets recorded from subjects walking around mazes, ticing patients,
> > and performance of three-ball juggling (in prep), repeated lifting of a
> > heavy box (in prep), etc. If I know my data are from laboratory recordings,
> > I would definitely use 5% trimming.
> > 
> > Makoto
> > 
> > On Thu, Dec 5, 2024 at 5:57 PM Joseph Dien via eeglablist <
> > eeglablist at sccn.ucsd.edu> wrote:
> > 
> > > I think it's an empirical question. If you're using a median, you are
> > > in effect relying on a single data point (albeit a median one) and it
> > > will have some amount of noise too. The question then is whether the
> > > noise you throw out with the rest of the points outweighs the signal
> > > that you throw out with the rest of the points. Or to put it another
> > > way, median is a great way to eliminate high amplitude noise, but not
> > > necessarily low amplitude noise. A graphic example that comes from my
> > > non-systematic observations of median averaging for ERPs (using my EP
> > > Toolkit), at least, is that it can give rise to sharp discontinuities in
> > > the waveform. I imagine a similar effect could happen in
> > > frequency-space. For robust ANOVAs, the developers (Wilcox, Keselman,
> > > Lix, Algina, et al) generally also recommend a 20% trim, but after
> > > mucking around with ERP data (again using my EP Toolkit), I came to the
> > > conclusion that a modest 5% trim was the better way to go (assuming n>=
> > > 20). Still unsystematic, though.
> > > 
> > > Definitely worth looking at!
> > > 
> > > Joe
> > > 
> > > On 12/5/24 11:30, Makoto Miyakoshi via eeglablist wrote:
> > > 
> > > > Hi
> > > > 
> > > > Cedric
> > > > 
> > > > 
> > > > ,
> > > > 
> > > > If you decide not to perform any signal processing for so called data
> > > > cleaning, do you think 20% trimming is trustworthy enough? I know it
> > > > is an empirical question. But in theory, median's breakdown point is 50%.
> > > > Since 50%>20%, why not?
> > > > 
> > > > Cyril once told me that if people assume Gaussian distributions
> > > > everywhere
> > > > anyway, why not use median? Good point.
> > > > 
> > > > Makoto
> > > > 
> > > > On Thu, Nov 28, 2024 at 7:50 AM Cedric Cannard via eeglablist <
> > > > eeglablist at sccn.ucsd.edu> wrote:
> > > > 
> > > > > > I have long thought that the averaging process in Welch's method should
> > > > > > be replaced with median because EEG recordings can contain pretty bad
> > > > > > outliers.
> > > > > 
> > > > > I recommend 20% trimmed mean :)
> > > > > 
> > > > > Cedric
> > > > > 
> > > > > On Monday, November 25th, 2024 at 12:25 PM, Makoto Miyakoshi via
> > > > > eeglablist eeglablist at sccn.ucsd.edu wrote:
> > > > > 
> > > > > > Hi Arno and all
> > > > > > 
> > > > > > Thanks for your input.
> > > > > > I recommend people posted here try that plugin and tell me what it
> > > > > > does.
> > > > > > eegfiltnew() and fir_filt() are close enough, if you just
> > > > > > need band-specific powers.
> > > > > > 
> > > > > > Makoto
> > > > > > 
> > > > > > On Sun, Nov 24, 2024 at 12:43 AM Arnaud Delorme via eeglablist <
> > > > > > eeglablist at sccn.ucsd.edu> wrote:
> > > > > > 
> > > > > > > For extracting brainwaves at specific frequencies, I would advise
> > > > > > > using
> > > > > > > the eegstats EEGLAB plugin.
> > > > > > > 
> > > > > > > Best wishes,
> > > > > > > 
> > > > > > > Arno
> > > > > > > 
> > > > > > > > On Nov 22, 2024, at 22:09, Meha Fatima via eeglablist <
> > > > > > > > eeglablist at sccn.ucsd.edu> wrote:
> > > > > > > > 
> > > > > > > > Hi. Thank you for this opportunity. I'd like to join in, too. Can you
> > > > > > > > please share the Zoom link?
> > > > > > > > 
> > > > > > > > On Fri, 22 Nov 2024 at 23:34, Makoto Miyakoshi via eeglablist <
> > > > > > > > eeglablist at sccn.ucsd.edu> wrote:
> > > > > > > > 
> > > > > > > > > Hi Devvarta,
> > > > > > > > > 
> > > > > > > > > > Could you please open this zoom session for others as well?
> > > > > > > > > > Sure, gather as many participants as you want and let them ask as
> > > > > > > > > > many
> > > > > > > > > > questions as necessary.
> > > > > > > > > > I'll be available after 7:00 am in New York time (EST) which is
> > > > > > > > > > 5:30 pm
> > > > > > > > > > in
> > > > > > > > > > Bangalore. I prefer it in the morning rather than the
> > > > > > > > > > evening/night.
> > > > > > > > > > I notice that there has been a large and constant need for
> > > > > > > > > > extracting
> > > > > > > > > > EEG
> > > > > > > > > > power using EEGLAB, which is only partially supported by GUI.
> > > > > > > > > > After our discussion, maybe I can summarize the requested
> > > > > > > > > > functionalities
> > > > > > > > > > into a solution in the form of EEGLAB plugin. It should be pretty
> > > > > > > > > > easy.
> > > > > > > > > > What I want to know is the type of output format you need most. Is
> > > > > > > > > > it
> > > > > > > > > > ascii
> > > > > > > > > > or Excel? Do you want a result in linear scale (microV^2) or in
> > > > > > > > > > dB? Do
> > > > > > > > > > you
> > > > > > > > > > want a mean power within a frequency band (with SD across sliding
> > > > > > > > > > windows)
> > > > > > > > > > or full freq bin resolution (which is already available in EEGLAB
> > > > > > > > > > spectopo)
> > > > > > > > > > etc..
> > > > > > > > > 
> > > > > > > > > I like the idea of robust PSD proposed by Melman and Victor (2016)
> > > > > > > > > so
> > > > > > > > > that
> > > > > > > > > you do not need to worry too much about data cleaning.
> > > > > > > > > https://urldefense.com/v3/__https://pmc.ncbi.nlm.nih.gov/articles/PMC4903894/__;!!Mih3wA!Hc_qvUxHKj94lakBpoAoDLDmyqzks4elWCUSSecAgiLfWsompHx83bc4_GJA8f1FYPh9Rceqb8iycYCC-UFVaSR4uL8$
> > > > > > > > > I have long thought that the averaging process in Welch's method
> > > > > > > > > should
> > > > > > > > > be
> > > > > > > > > replaced with median because EEG recordings can contain pretty bad
> > > > > > > > > outliers. At least, outcomes from Welch's method by mean and
> > > > > > > > > median can
> > > > > > > > > be
> > > > > > > > > compared to evaluate the window-mean PSD's deviation from the
> > > > > > > > > window-median
> > > > > > > > > PSD.
> > > > > > > > > 
> > > > > > > > > Grace, if this selection of time does not work for you, we can set
> > > > > > > > > a
> > > > > > > > > separate Zoom meeting. Tell me what time zone you are in.
> > > > > > > > > 
> > > > > > > > > Makoto
> > > > > > > > > 
> > > > > > > > > On Fri, Nov 22, 2024 at 3:42 AM Devvarta Kumar <
> > > > > > > > > devvarta.k at nimhans.ac.in>
> > > > > > > > > wrote:
> > > > > > > > > 
> > > > > > > > > > Hi Mokoto
> > > > > > > > > > I assume many of us would be interested in understanding the
> > > > > > > > > > process of
> > > > > > > > > > calculating power in different bands in raw EEG. Could you
> > > > > > > > > > please open
> > > > > > > > > > this
> > > > > > > > > > zoom session for others as well?
> > > > > > > > > > Regards
> > > > > > > > > > 
> > > > > > > > > > Devvarta Kumar, Ph.D.
> > > > > > > > > > 
> > > > > > > > > > Professor
> > > > > > > > > > Department of Clinical Psychology
> > > > > > > > > > M. V. Govindaswamy Building
> > > > > > > > > > National Institute of Mental Health and Neurosciences,
> > > > > > > > > > Hosur Road,
> > > > > > > > > > Bangalore,
> > > > > > > > > > Karnataka-560029
> > > > > > > > > > India
> > > > > > > > > > 
> > > > > > > > > > Ph: +91-80-26995188
> > > > > > > > > > https://urldefense.com/v3/__http://www.nimhans.ac.in/users/dr-devvarta-kumar__;!!Mih3wA!E8Vsq9Vu8Q4ZAl8UIb5elP9m72UAtG-_h15RsH4N7qURhjnEctf1NKia1NpntFgQlL_UtZbHh_sQjnDjaxBj7B3Oazg$
> > > > > > > > > > ResearchGate:
> > > > > > > > > > https://urldefense.com/v3/__https://www.researchgate.net/profile/Devvarta_Kumar/contributions__;!!Mih3wA!E8Vsq9Vu8Q4ZAl8UIb5elP9m72UAtG-_h15RsH4N7qURhjnEctf1NKia1NpntFgQlL_UtZbHh_sQjnDjaxBj8leJO1o$
> > > > > > > > > > ----- Original Message -----
> > > > > > > > > > From: eeglablist at sccn.ucsd.edu
> > > > > > > > > > To: eeglablist at sccn.ucsd.edu
> > > > > > > > > > Sent: Friday, November 22, 2024 1:51:05 AM
> > > > > > > > > > Subject: Re: [Eeglablist] EEGLAB Orientation
> > > > > > > > > > 
> > > > > > > > > > Hi Grace,
> > > > > > > > > > 
> > > > > > > > > > If your group wants to Zoom with me, I'll be available. Which
> > > > > > > > > > time zone
> > > > > > > > > > are
> > > > > > > > > > you in?
> > > > > > > > > > 
> > > > > > > > > > Makoto
> > > > > > > > > > 
> > > > > > > > > > On Thu, Nov 21, 2024 at 2:50 PM Grace Bang via eeglablist <
> > > > > > > > > > eeglablist at sccn.ucsd.edu> wrote:
> > > > > > > > > > 
> > > > > > > > > > > Hello!
> > > > > > > > > > > 
> > > > > > > > > > > We are 3rd year medical students trying to conduct research
> > > > > > > > > > > where we
> > > > > > > > > > > collect brain wave activity data using the CGX Patch 2-channel
> > > > > > > > > > > EEG.
> > > > > > > > > > > We've
> > > > > > > > > > > been having a lot of issues trying to navigate how to use
> > > > > > > > > > > EEGLAB to
> > > > > > > > > > > process
> > > > > > > > > > > the data. We just want to be able to quantify the number of
> > > > > > > > > > > alpha and
> > > > > > > > > > > beta
> > > > > > > > > > > waves present in the raw data. The resources we've read online
> > > > > > > > > > > have
> > > > > > > > > > > been
> > > > > > > > > > > somewhat helpful, but we are looking for someone who can jump
> > > > > > > > > > > on a
> > > > > > > > > > > Zoom/phone call to help fill some gaps of knowledge. Neither
> > > > > > > > > > > of us
> > > > > > > > > > > have a
> > > > > > > > > > > background in more complicated softwares/coding/EEGs so any
> > > > > > > > > > > guidance
> > > > > > > > > > > would
> > > > > > > > > > > be greatly appreciated!
> > > > > > > > > > > 
> > > > > > > > > > > - Grace Bang, OMS-3
> > > > > > > > > > > _______________________________________________
> > > > > > > > > > > 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
> > > > > > > > > > > --
> > > > > > > > 
> > > > > > > > * Dr. Meha Fatima BSMT, Ph.D.*
> > > > > > > > Assistant Professor
> > > > > > > > Dow Institute of Medical
> > > > > > > > Technology
> > > > > > > > Dow University of Health Sciences
> > > > > > > > mehafatima1 at gmail.com
> > > > > > > > mehafatima1 at gmail.com.
> > > > > > > > _______________________________________________
> > > > > > > > --------------------------------------------------------------------------------
> 
> 
> Joseph Dien, PhD
> Senior Research Scientist
> Department of Human Development and Quantitative Methodology
> University of Maryland, College Park
> https://urldefense.com/v3/__http://joedien.com__;!!Mih3wA!F0eaU-cVQEnR4py5rJjda2ULVa9FGrPCPmsQtgXL7IUtxwwzJLgmIKRtIG-xy8RTjfoVhALfeyt1ZVZx0rc$
> 
> _______________________________________________
> 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