I'm trying to estimate the mean power over time in different bands for different labels.

I've tried the two following approaches:

1) Running mne.minimum_norm.source_band_induced_power on the epochs, (which is eating almost all the memory of our RAM monster)

2) First calculating the labels time course from the stc, then calculating the power using scipy.signal.welch

I've noticed, that for example for one of the labels and one of the bands, Welch gives me 1.886, wherein the first approach, the stc.data.mean(0) is between 45,260 and 145,023.

I'm trying to estimate the mean power over time in different bands for different labels.

I've tried the two following approaches:

1) Running mne.minimum_norm.source_band_induced_power on the epochs, (which is eating almost all the memory of our RAM monster)

2) First calculating the labels time course from the stc, then calculating the power using scipy.signal.welch

I've noticed, that for example for one of the labels and one of the bands, Welch gives me 1.886, wherein the first approach, the stc.data.mean(0) is between 45,260 and 145,023.

What am I'm missing?

I am not sure it's the reason but source_band_induced_power using
Morlet wavelets while welch
is using an averaged periodogram with fixed window size.

Note that source_band_induced_power has not much been improved since
it was written
7 years ago...