Psd Vs magnitude values

Discussions
,
1

Hi
I'm trying to perform a frequency analysis, but I'm not sure about the values:

I computed Welch psd for many eeg files, when I try to run a T statistical Test I get an error message suggesting not to use power values but magnitude values, and then I got another error message when I try to view the T test touput:

BST> Average corrected p-threshold: 0.00263158 (FDR, Ntests=19)
Warning: Cannot determine t-test side

In process_extract_pthresh>Compute (line 378)
In process_extract_pthresh (line 28)
In bst_memory>LoadTimefreqFile (line 1554)
In bst_memory (line 72)
In view_spectrum (line 72)
In tree_callbacks>@(h,ev)view_spectrum(filenameRelative,'Spectrum') (line 2067)

I used to work with the power values or normalize it with the log transform, but I don't understand what the" Magnitude " means and what is for, I extract the magnitude values from a file and I compared with the power values, but it does not seem to be just the squared value:

PSD:

Freq delta
Fp1 5.9245E-12

Magnitude:

Freq delta
Fp1 2.43403E-06

thanks in advance!

2

I computed Welch psd for many eeg files, when I try to run a T statistical Test I get an error message suggesting not to use power values but magnitude values

In general, parametric tests is not adapted to compare power values, because they are not normally distributed. Use non-parametric tests instead. With a permutation test, you can test any type of values, no matter what their distributions are. See the Statistics tutorial, especially the part testing for normality of the distributions:
https://neuroimage.usc.edu/brainstorm/Tutorials/Statistics

Warning: Cannot determine t-test side

This is just a minor warning related with the display of the values. It does not adjust the colors in the best possible way, but it doesn't matter much, the displays are still correct.

I extract the magnitude values from a file and I compared with the power values, but it does not seem to be just the squared value

I'm not sure what you mean, your example shows that one is exactly the square of the other:
2.43403E-06 * 2.43403E-06 = 5.9245e-12

3

I see :sweat_smile:

Thanks!