Understanding the unit and outputs of compute_source_psd_epochs()

💬 Support & Discussions
1

Hi everyone,

Recently, with a collaborator / student the question came up how to understand the output of our source-space multitaper function compute_source_psd_epochs

The example on our website does not provide y-axis labels and the documentation is quite sparse.

What would be important to understand is how to describe interpret the output, e.g., when composing a methods section, e.g., is the output log power? What are the units?

Based on the source code my understanding, this depends on the methods param (e.g. some F-like statistic if dSPM vs current density for MNE) and that the output is not log power.

Any thoughts on this? Should we open an issue on Github?

Btw. I notice that the viz output in examples like Computing various MNE solutions — MNE 0.24.1 documentation for MNE is AU but that is not true no? It’s physical units in that case.

Any feedback would be welcome.

1 Like
2

well putting a unit is not so simple as it depends on the inverse operator / method you use.
Here the examples uses dSPM which has no unit. But with MNE you would get nAm^2 / Hz

Alex

3

Thanks for the reply! Yes, that was my intuition and the background of the question. So do I read you correctly that this also implies that the multitaper procedure as implemented in this function does not apply explicitly any logarithmic transform (as in dB units)? On the flip side, do we get some implicit log power when using dSPM or LORETA as inverse methods? Literally asking for a friend :smiley:
We should probably push the results of this discussion into the MNE documentation.

1 Like