[Eeglablist] **data length

Tarik S Bel-Bahar tarikbelbahar at gmail.com
Thu Jul 13 08:10:37 PDT 2017

​Hello Jaleh,

Please go/google with the published recommendations on the eeglab tutorial,
and based on Makoto's past recommendations on the eeglablist.

Basically, the more channels and higher sampling rate you have, the longer
your time periods should be. It is also important to keep in mind that one
wants to feed enough data to ICA so that it can adequately separate sources.

Basically, for a great/valid ICA decomposition, it is better to have more
(rather than less) data (e.g., ~30 minutes or more) for each condition,
especially when you have high-density eeg with 64 or more channels, and
when you have high sampling rate.

In your case, you may want to reduce the total number of channels, or
downsample the data, in order to better meet ICA's requirements.
You may also want to use a PCA reduction. Overall, such steps should help
get valid ICA decompositions from data that are relatively short (in terms
of time).

Please note that researcher preferences and biases vary greatly in the EEG
field, do not expect there to be agreement on a range of issues. It is a
good path to follow guidelines from top researchers, and methods from
articles in top journals.

Note that some researchers use (and publish successfully on) 2 or 5 minute
periods for ICA, and consider that enough. Such short periods often can
result in relatively clear artifact ICs. You can test this for yourself.
You can also test the effects of having longer versus shorter periods.
However, the validity of ICA results is likely affected by using using
short periods, and the decomposition may not show many valid neural ICs.

Your best best is to review past studies using ICA and note how many
channels and how many time points entered into the ICA, and what the
results were.


On Wed, Jul 12, 2017 at 1:34 AM, Jaleh Mohammad alipoor <sugmad973 at yahoo.com
> wrote:

> Hello,
> l want to know what is the suitable length of eegdata for ICA to work
> properly. l know that there is a formula as k*n^2 but adding to the
> confusion is the fact that different numbers has mentioned for 'k' !! like
> 30, 25 and even 2!
> for example l want to know if l run ICA for a data of 13000 data point
> while my channel numbers are 32, will  ICA work correctly or not?
> Best regards,
> Zhaleh
> _______________________________________________
> 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...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20170713/82ac9f27/attachment.html>

More information about the eeglablist mailing list