Removing ERP response before instantaneous PLV computation

Dear Brainstorm community,

I have a question regarding the impact of removing or not the ERP/ERF before computing source connectivity with the instantaneous phase locking value.

At present the option of removing the ERP is not available in Brainstorm, so I cannot check it myself with data. So I was wondering if you could provide some explanation of how much keeping the ERP could bias the estimation of the time-varying connectivity, and if so how is it possible to minimize for this bias?

Unfortunately I was not able to find any paper that directly addressed this issue :man_shrugging:


@Sylvain @Samiee @hossein27en @pantazis?

Hi @danielemarinazzo I have seen you super active in the last days together with @Francois in the new implementation of the phase-locking value, and I was wondering if you can help me with this issue that is making me struggling a lot :sweat_smile:

Thank you very much,

The fact that I try to keep myself busy coding is maybe that I don't want to think about stuff :slight_smile:

Now, there is no general answer to the question whether one should remove the ERP or not. It heavily depends on your experiment, and your hypotheses.

For example, if you want to relate it with behavioral outcomes, in particular binary ones, then the phase of the ongoing activity is probably relevant, see for example

A less controversial thing is that if you remove the ERP this should be done on a single trial basis, otherwise you will introduce dependencies, see

as this often happens, this is more true for autoregressive models rather than zero lag ones.

Now this said, it would certainly be good to have all the options at hand in a toolbox.

Hi @danielemarinazzo thank you very much for your response and sorry for forcing you out of your coding bubble :sweat_smile:

I actually happened to read in the last weeks some of the papers you mentioned, like

so if my understanding is correct there's no "safe" way to study task-based dynamic connectivity with M-EEG if the task elicits an ERP/ERF?

Nothing is perfect or foolproof. We are "safe" to the extent to which our experiments are well designed and we know the methods that we apply.
The papers linked above point to problems but also to some solutions. Of course it's difficult to have a perfect and exclusive link between a manipulation and the measured outcome, but that's a general thing, not specific to time-varying phase-based connectivity.

1 Like

You are 100% right Daniele.

Thank you very much for your feedback!