[Eeglablist] (Too) long of an explanation and questions at the end
mmiyakoshi at ucsd.edu
Tue Feb 18 19:19:53 PST 2014
I wish you good luck in launching new experiments in the new field! Here
are my advice to perform a good experiment with lipstick piggies.
1. Choose a robust paradigm.
2. For event-related potential paradigm, try to have as many trials as
possible. 100 would be good to start.
3. For this reason, minimize the conditions to compare. The best is 1x2
design (within-subject, 2 conditions) so that you can have more trials.
4. Record as long data as possible. 7-8 min rest before and after the
taking a cap* helps your ICA decomposition during the task period.
5. Design the task so that you can study your phenomenon of interest even
with behavioral measurement only. Behaviorally observable difference *must
be* reflected in EEG.
Don't forget that robust paradigm, many trials, and long recording time are
the best solutions for good data, and signal processing does not help as
2014-02-16 23:53 GMT-08:00 Sharon Jalene <sharonjalene at gmail.com>:
> Lipstick on a pig, eh? I can't stop laughing. Hopefully I can get enough
> interest with this project to get a better looking pig. Thanks for the link!
> On Sun, Feb 16, 2014 at 11:27 PM, Tom Campbell <tom_campbell75 at hotmail.com
> > wrote:
>> The most serious noise issue concerns the electrodes. So the electronics
>> are rather good at putting lipstick on a pig! The device has built-in
>> filters, though additional offline filtering is an option. People use this
>> device for BCI without offline filtering. It is possible to customise with
>> better electrodes:
>> These may be of interest:
>> Debener, S., Minow, F., Emkes, R., Gandras, K. & de Vos, M. How about
>> taking a low-cost, small, and wireless eeg for a walk?.
>> *Psychophysiology* *49* 1617-1621 doi:
>> http://dx.doi.org/10.1111/j.1469-8986.2012.01471.x ( 2012)
>> From: arno at salk.edu
>> Date: Fri, 14 Feb 2014 20:44:45 -0800
>> To: sharonjalene at gmail.com
>> CC: eeglablist at sccn.ucsd.edu
>> Subject: Re: [Eeglablist] (Too) long of an explanation and questions at
>> the end
>> *1. **Should I epoch the data before statistical analyses even though
>> I am looking at the differences between the Grand Average - within
>> Yes, you need to extract data epochs to compute the grand average ERP. If
>> you are computing the spectrum (based on the rest of your message), then
>> epoch extraction is not necessary.
>> *2. *I am HP filtering at 1Hz and using Cleanline, automatic artifact
>> rejection, visual artifact rejection. I have played with ICA component
>> rejection, but am not sure I have enough good data to do so. (I recorded
>> pilot data on one subject - I have 3 minutes of data for each condition and
>> 14 electrodes). *Is there an optimal filtering method since I am using
>> the EMOTIV?*
>> This is a very noisy system. To extract meaningful ICA components, the
>> more filtering the better.
>> *a. *One of my committee said I should NOT filter the data or
>> remove artifacts since I am just looking for the Grand Average of certain
>> freq bands. Although I see his logic, I believe I should still remove
>> spurious data so that the mean is not skewed by muscle and other artifacts...
>> Yes, if you have strong drifts in your data or DC shifts, you will need
>> to filter. Otherwise, it will ruin the whole spectrum.
>> *3. *I was able to create a STUDY. I saved separate data sets with
>> the electrodes of interest. I was able to run power spectrum channel stats
>> as a 3 (conditions) X 2(Left and Right) ANOVA. I saw the t-tests
>> between conditions and the interaction.
>> *a. *Is the power spectrum giving me the average of the specified
>> freq range?
>> *b. *If I epoch the data, can I use Bootstrapping instead of
>> parametric or permutation?
>> No, this is independent.
>> *c. *How can I get the actual values, including the p values, of the
>> stats? I see the option to show a table with statistics containing the
>> median, mean, mode, etc., but it rarely populates and is inconsistent. IS
>> that statcond - on?
>> Look at the history and add output to the function that plot the STUDY
>> spectrum (the function name is std_spec). This function will return the
>> statistics (which are computed by statcond).
>> *d. *If I show results from a bootsrap, what statistical test was
>> used to obtain the results?
>> The t-value (paired or unpaired) or the F-value is computed. The
>> bootstrap procedure randomly samples the data and allow EEGLAB to obtain
>> confidence intervals for these values under the null hypothesis.
>> *e. *Would I be better off using a ratio of Grand Average
>> subtraction of the control from the FOCUS conditions. or pwelch in MATLAB
>> -then dropping the data into SPSS for some non parametric analysis
>> (WIlcoxon rank ordered)?
>> Sure. You should try what you are comfortable with. The results should be
>> _______________________________________________ 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
> Sharon Jalene
> office: 702-966-3010
> mobile: 303-908-8441
> *If one dream should fall and break into a thousand pieces, never be
> afraid to pick one of those pieces *
> *and begin again.*
> *Flavia Weedn*
> 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
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