Relative vs Absolute Bandpowers

[stellarpower]

Salut!

I have been wondering about whether I should be using relative or absolute bandpowers in my own implementation of protocols using some other software.

As far as I can see from the FFT and feedback calculation, it seems that neurofeedbacklab is using absolute bandpowers by default (please correct me if I have missed something here).

If I google this, I see some mention of both approaches, but nothing specifically clarifying any testing of this empirically in neurofeedback sessions. Has neurofeedbacklab ever been tried using relative bandpowers? Are there reasons why one approach is better than the other, is it something that would be advantageous to add into the code?

In my own data, I notice that the power of inhibit and reward channels are correlated - especially if there are artefacts from e.g. moving, we see an increase in power in the signal across all frequencies and the DC bias. This would suggest relative bandpower might prove more effective in trainnig. But on the other hand, we would need to be careful about what we compare it to. If artefacts are entering into the protocol (ideally they would be filtered, but this is not perfect), then we risk taking the noise and outliers featuring in the power across all frequencies and distorting the reward band from this. Personally, the first thing I want to try is taking a select band of relatively safe frequencies (maybe 3-20Hz) and taking this as our definition of the total bandpower, and see how stable that is. But I wanted to ask here to see if there is any experience of testing that and if it's desireable or there are some contra-indications.

Thanks a lot!