spectra plot of the covariance explanation

Support & Discussions
1

Hi,

I would like an explanation to what the below plot is actually depicting

sphx_glr_30_mne_dspm_loreta_002
sphx_glr_30_mne_dspm_loreta_002760×370 24.8 KB

This is the spectral plot of the noise covariance.

Does this show the noise level at each channel and why the rank?

Thanks

2

These are SVD plots, not spectral plots. They depict singular values plotted over eigenvalue indices.

What you will want to look for here is when a “cliff” occurs, i.e., at which eigenvalue index the singular value suddenly drops sharply. That index gives you an estimate of the rank of the data, because a small singular value basically means that there’s no additional information in this feature.

These types of plots are therefore very important diagnostic visualizations.

Best wishes,
Richard

3

Thanks for alot. I understand it better now. Actually in the documentation, the output is SVD spectral plot of covariance.

4

can you provide a link to the documentation page where it says this? We should correct it, as these are not “spectra”

5

https://mne.tools/stable/generated/mne.viz.plot_cov.html#mne.viz.plot_cov

that’s the link

Thanks

6

thanks; will be fixed in fix misleading docstring return description for covariance SVD plot by drammock · Pull Request #12359 · mne-tools/mne-python · GitHub