[Eeglablist] using 'psd' or 'fft' in std_spec creates 60dB difference

[Seyed Yahya Shirazi]

Hi,
I posted this issue on EEGLAB GitHub a few days ago. Sorry for
cross-posting. I use "std_precomp" to create the power spectra of the ICA
components in a source cluster. I tried computing the spectra using `psd` or
`fft` and results have almost 60 dB difference (it is not exactly dB, but
more 10Log(power), since it does not have a reference power). From
experience, and also the time-frequency data, I know that the results of the
`fft` mode might be correct.

Also, the `fft` results appeared to be very noisy. Do you have any comments
for adjusting parameters to get smoother results?

A friend of mine at another university also observed this discrepancy using
a totally different dataset but a similar (yet independent) pipeline.

the full script for the computing spectra is as follows:

% PSD
[f_STUDY, f_ALLEEG] = std_precomp(f_STUDY, f_ALLEEG, 'components',
'recompute' ,'on','spec','on',...
                'specparams', {'specmode', 'psd', 'logtrials', 'off',
'freqrange',[3 80], 'freqfac', 10}, ...
                'recompute', 'on', 'savetrials', 'on');

% FFT
[f_STUDY, f_ALLEEG] = std_precomp(f_STUDY, f_ALLEEG, 'components',
'recompute' ,'on','spec','on',...
                'specparams', {'specmode', 'fft', 'logtrials', 'off',
'freqrange',[3 80]}, ...
                'recompute', 'on', 'savetrials', 'on');


BTW, changing logtrials and/or freqfac values do not seem to make any
change. I am using EEGLAB 2019.0 and Matalb R2018b.

The report together with an image of the difference between the two modes of
computing spectra are available at https://urldefense.com/v3/__https://github.com/sccn/eeglab/issues/172__;!!Mih3wA!SqkCLK1Bq7r6WLhrG5pqnP4M1UqCWdc0qyr8uA3E-vs-Gt0p5Mhqwtj2Sh9EGR1CT954Bw$ 

Thanks
Seyed

--
Seyed Yahya Shirazi
GRA, BRaIN Lab
University of Central Florida