<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
<title></title>
</head>
<body bgcolor="#ffffff" text="#000000">
Stefan Scherbaum wrote:
<blockquote cite="mid60e9cc4e050729025030fe406a@mail.gmail.com"
type="cite">
<pre wrap="">I have a question about converting sampling rates in EEGlab, hoping
that you can help me:
I recorded a file with 1000 Hz and an analog filter onset at 250 Hz.
Later, I downsampled the data with Brain Vision Analyzer to 250 Hz,
using a softare filter at 112.5 Hz with 24db damping. This was done to
make the import into EEGLab easier.
To check the results of the downsampling, I imported one channel of
the original data to EEGlab and downsampled it to 250 Hz after
filtering at 112.5 Hz.
Comparing this channel with the channel converted with BVA (and
imported to EEGLab), I realized differences that I can not explain and
that I hope you might possibly have an explanation for.
1. The length of the data converted by eeglab (724581) is not 1/4 of
the original length (2898320) - the BVA file has exactly 1/4 of the
length (724580).
</pre>
</blockquote>
I created a dataset with 2898320 data points, filtered and resampled
the data and obtained exactly 724580 data points. For the difference in
data points, my simpliest explanation would be that the sampling rate
imported in EEGLAB (1000 Hz) is not a round number (1000.0001 Hz for
instance). This number may be rounded in some graphic interfaces. You
can check that by typing EEG.srate. If this is not the case, then I
cannot explain this discrepency since I do not observe it.<br>
<blockquote cite="mid60e9cc4e050729025030fe406a@mail.gmail.com"
type="cite">
<pre wrap="">2. There are differences in the curves - I attached a screenshot
(timescale 0.25): channel 1 is the channel converted by EEGLab (-1
points), channel 2 is the data converted by BVA and channel 3 shows
the difference.</pre>
</blockquote>
The two datasets may look slightly different because different filters
have
been used. I do not know which filter BVA uses but it is probably not
linear (24 dB damping seems beyond the capabilities of a linear
filter). You may try
using the IIR non-linear filter plugin in EEGLAB and see if the result
is
closer to what is returned by BVA.
<blockquote cite="mid60e9cc4e050729025030fe406a@mail.gmail.com"
type="cite">
<pre wrap="">Do you have an idea about all this? I am a little bit confused which
data I should use for further analysis - do you have an idea how to
choose the right dataset to continue?
</pre>
</blockquote>
As far as which dataset you should use, I think using either dataset is
fine as long as you do not forget to mention which filter type you used
(and which software).<br>
<br>
Best regards,<br>
<br>
Arno<br>
<blockquote cite="mid60e9cc4e050729025030fe406a@mail.gmail.com"
type="cite">
<pre wrap="">Thank you very much!
Best regards,
Stefan</pre>
</blockquote>
<br>
<div class="moz-signature">-- <br>
<br>
<b><font face="Arial,Helvetica"> <a
href="http://www.sccn.ucsd.edu/%7Earno" target="_blanc">Arnaud Delorme</a>,
Ph.D.</font></b>
, SCCN, UCSD, San Diego, USA
<br>
<br>
<table border="0" cellpadding="0" cellspacing="0" width="80%">
<tbody>
<tr>
<td><b>SCCN main page:</b></td>
<td><a href="http://www.sccn.ucsd.edu/" target="_blanc">www.sccn.ucsd.edu</a></td>
</tr>
<tr>
<td><b>EEGLAB main page:</b></td>
<td><a href="http://www.sccn.ucsd.edu/eeglab" target="_blanc">www.sccn.ucsd.edu/eeglab/</a></td>
</tr>
<tr>
<td><b>EEGLAB documentation:</b></td>
<td><a href="http://www.sccn.ucsd.edu/eeglab/eeglabdocs.html"
target="_blanc">www.sccn.ucsd.edu/eeglab/eeglabdocs.html</a></td>
</tr>
<tr>
<td><b>EEGLAB FAQ page:</b></td>
<td><a href="http://www.sccn.ucsd.edu/eeglab/eeglabfaq.html"
target="_blanc">www.sccn.ucsd.edu/eeglab/eeglabfaq.html</a></td>
</tr>
<tr>
<td><b>EEGLAB mailing lists:</b></td>
<td><a href="http://www.sccn.ucsd.edu/eeglab/eeglabmail.html"
target="_blanc">www.sccn.ucsd.edu/eeglab/eeglabmail.html</a></td>
</tr>
</tbody>
</table>
</div>
</body>
</html>