[Eeglablist] High pass filtering and ICA

[Gilbert, David G]

For the application of 0.5Hz high-pass to the 0.1Hz file in ERP analyses, can the files be epoched when the file 2 ICAs are applied to the file 1 (0.1 Hz HP)? Is there a video showing how this is done? I keep getting an error using epoched data.
Thanks in advance,
David

________________________________
From: Gilbert, David G
Sent: Friday, September 27, 2024 3:43 PM
To: Gilbert, David G <dgilbert at siu.edu>; Gunn, Matthew <mgunn at uic.edu>
Subject: High pass filtering and ICA

eeglablist<eeglablist-bounces at sccn.ucsd.edu> on behalf of  Makoto Miyakoshi via eeglablist<eeglablist at sccn.ucsd.edu>
To:
​
eeglablist at sccn.ucsd.edu
​Fri 9/27/2024 3:28 PM
[EXTERNAL EMAIL ALERT]: Verify sender before opening links or attachments.

Hi Cedric,

> But I haven't found any perfect solution personally.

The perfect solution, if you want, is to process the EEG data in two
different ways.

1. High-pass filter at 1-2 Hz + ICA; This preprocessing is optimum for
ICA + time-frequency decomposition above the cutoff frequency of the HPF,
but not for time-domain averaged ERP because some people say
this 'distorts' the waveforms.
2. High-pass filter at < 0.5 Hz + ICA; This one is optimum for time-domain
averaged ERP but not so for ICA + time-frequency analysis.

Who said we can't do it, so why not?
I think using the above approach is more straightforward than worrying
about whether to use the HPF trick for ICA.

An important confirmation is to compare the ICA results between 1 and 2
(before that, repeat ICA several times on the same data to see how
reproducible the results are!) If they are practically the same, there is
no need to wonder about the choice.

Makoto