<div dir="ltr"><div class="gmail_default" style="color:rgb(51,51,153)">Hello Yamil, some notes below, hope you find them of use. cheers!</div><div class="gmail_default" style="color:rgb(51,51,153)"><br></div><div class="gmail_default" style="color:rgb(51,51,153)">see Groppe et al on removing baseline for better ICs, and search past eeglablist discussions on Google with your keywords, as similar questions have come up before.</div><div class="gmail_default" style="color:rgb(51,51,153)"><br></div><div class="gmail_default" style="color:rgb(51,51,153)">most would say it's a good thing to remove at least low-frequency drifts. most would say it's okay to do a 1hz highpass, and/or "remove baseline for each channel in the continuous data" for better ICs. be wary of a 1hz highpass if you're doing erps.</div><div class="gmail_default" style="color:rgb(51,51,153)"><br></div><div class="gmail_default" style="color:rgb(51,51,153)">yes it does not matter if you give ICA trials or continuous data, spatial ICA as implemented in eeglab does not care about time for finding stable scalp/source maps. It does matter what kind of data you give it, but that's another discussion.</div><div class="gmail_default" style="color:rgb(51,51,153)"><br></div><div class="gmail_default" style="color:rgb(51,51,153)">I guess one would not include much more than a 1000 ms or 500 ms baseline. baselines are usually included for ICA of multiple trials/epochs/segments, especially in erp protocols. Check the literature if people have done much baseline-subtraction for epochs before ICA.</div><div class="gmail_default" style="color:rgb(51,51,153)"><br></div><div class="gmail_default" style="color:rgb(51,51,153)">You might be interested in recent single-trial denoising and metrics work, check on Google Scholar.</div><div class="gmail_default" style="color:rgb(51,51,153)"><br></div><div class="gmail_default" style="color:rgb(51,51,153)"><br></div><div class="gmail_default" style="color:rgb(51,51,153)"><br></div><div class="gmail_default" style="color:rgb(51,51,153)"><br></div><div class="gmail_default" style="color:rgb(51,51,153)"><br></div><div class="gmail_default" style="color:rgb(51,51,153)"></div></div><div class="gmail_extra"><br><div class="gmail_quote">On Wed, Oct 21, 2015 at 7:50 AM, Yamil Vidal Dos Santos <span dir="ltr"><<a href="mailto:hvidaldossantos@gmail.com" target="_blank">hvidaldossantos@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">Hi all,<div>I have a question regarding getting data ready for ICA.</div><div>As one factor that affects the quality of an ICA decomposition is the stationarity of the data, I decided to segment and detrend my data before running ICA. But I have read that if one would run ICA on segmented data, one should have a long enough baseline and/or should not remove baseline.</div><div>This sounds strange to me, because as far as I know, ICA is not concerned about time. Furthermore, data is whitened before running ICA. Doesn't this imply a baseline removal?</div><div><br></div><div>My concrete question is about the usefulness of detrending to improve data stationarity before ICA, but any clarifications about how to improve the chances of getting a good ICA decomposition will be appreciated.</div><div>Thanks,</div><div>Yamil</div></div>
<br>_______________________________________________<br>
Eeglablist page: <a href="http://sccn.ucsd.edu/eeglab/eeglabmail.html" rel="noreferrer" target="_blank">http://sccn.ucsd.edu/eeglab/eeglabmail.html</a><br>
To unsubscribe, send an empty email to <a href="mailto:eeglablist-unsubscribe@sccn.ucsd.edu">eeglablist-unsubscribe@sccn.ucsd.edu</a><br>
For digest mode, send an email with the subject "set digest mime" to <a href="mailto:eeglablist-request@sccn.ucsd.edu">eeglablist-request@sccn.ucsd.edu</a><br></blockquote></div><br></div>