[Eeglablist] Preprocessing - Doubts regarding filtering and removing line noise

[Makoto Miyakoshi]

Dear Katia,

> a) If my low pass filter is set to 45 Hz, do I still need to try and
remove
the line noise (clean line)?  As far as I understand line noise is usually
at a higher frequency, so does it mean that my filter will have taken care
of it already, or am I confusing things?

If your power line frequency is at 60 Hz, No.
If 50 Hz, you may want to design your filter such that it has good
suppression at 50 Hz--then no need to deal with line frequency artifact.
See the following page to learn how to design a good filter. It is quite
easy!
https://sccn.ucsd.edu/wiki/Firfilt_FAQ

> b) Is it correct that it is good practice to apply a low and high filter
in
two separate steps? I am sure that I read that somewhere, but I cannot find
it anymore and, assuming I haven't dreamt it, I cannot remember what the
reasoning was.

Often yes.
If you apply 1-Hz transition band width for 1-Hz high-pass filter, the
transition band width is 0.5-1.5 Hz.
However, if you do the same for 45 Hz, it will be 44.5-45.5 Hz, which is
quite steep.

Do you still wonder why? Compare the following examples. They are both
1-octave range.
1-2Hz
40-80Hz

> c) Also, more curious than anything: the first thing I did in the
preprocessing (after assigning channel locations) was to downsample to
250Hz. If I then filter out everything above 45Hz anyway, is downsampling
even necessary?

Good point!
When downsampling the signal to 250 Hz, the antialiasing filter (that is
automatically taken care of in using Matlab) is usually at around 250/2 =
125 Hz.
However, sometimes people apply 250/3, 250/4, and even 250/5 Hz. This is
called oversampling.
https://urldefense.com/v3/__https://en.wikipedia.org/wiki/Oversampling__;!!Mih3wA!RWcRClOnux3iEU3Ouy8C9bLsI92LZQmtqpCZchPqxFBh3yZTWYjFN6r6ITRb4GIGUohQYA$ 
What's so good about oversampling? You will certainly have better phase
resolution near the highest frequency range available. Otherwise... a good
placebo effect?

Makoto



On Thu, Jul 9, 2020 at 10:23 AM Katarzyna Dudzikowska <
k.a.dudzikowska at gmail.com> wrote:

> Dear all,
>
> I am working on preprocessing data with EEGLAB with the end goal of running
> a frontal asymmetry analysis. As I am trying to settle on proper
> preprocessing steps, I have a few issues/doubts and I would love to get
> some input from the more knowledgeable crowd (and I am certain you are as
> this is my first time working with EEG data :) )
>
> As I will be looking at alpha frequency, I judged that filtering between
> 0.5 and 45Hz should work for me, but this begets more questions
> (obviously).
> a) If my low pass filter is set to 45 Hz, do I still need to try and remove
> the line noise (clean line)?  As far as I understand line noise is usually
> at a higher frequency, so does it mean that my filter will have taken care
> of it already, or am I confusing things?
> b) Is it correct that it is good practice to apply a low and high filter in
> two separate steps? I am sure that I read that somewhere, but I cannot find
> it anymore and, assuming I haven't dreamt it, I cannot remember what the
> reasoning was.
> c) Also, more curious than anything: the first thing I did in the
> preprocessing (after assigning channel locations) was to downsample to
> 250Hz. If I then filter out everything above 45Hz anyway, is downsampling
> even necessary?
>
> I will be extremely grateful for your input.
>
> All the best,
> Katia Dudzikowska
>
> P.S. I have some other questions, but as they regard other preprocessing
> steps I figured I would ask them separately, for the sake of future
> searches, hope it's ok.
> _______________________________________________
> 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
>