[Eeglablist] Issues in effective connectivity

[Daniele Marinazzo]

Dear Danish

First, the usual distinction between map and territory (i.e. the target
properties we want to investigate and the methods used to measure them).
Personally I would refer as "effective connectivity" to methods involving a
generative model of brain activity, e.g. Dynamic Causal Models. Also in
that case it's easier to speak of "ground truth", even if of course you
cannot use the same method to generate the ground truth and to validate it.
DCM works on fitting the whole spectrum rather than individual frequencies,
even though individual frequency contributions would have an overall impact.

If on the other hand we have methods based on prediction, such as Granger
causality, it exists in the frequency domain (but without fitting two
separate models, which is problematic as Stokes and Purdon polularized
https://urldefense.com/v3/__https://www.pnas.org/doi/10.1073/pnas.1809324115__;!!Mih3wA!DRA7wpoO7Sj5UwTfCHa12-H8Fj6hZlTbZPCzDs2cqW8NRNaOiwfO8pSAzRjak9_P7oepAyKwRY-b0Y1HkmXPGHNgdojYjgHL$  even if they ignored that
several solutions existed already).
In this case though, "causality" or "connectivity" is in terms of effects,
not of mechanisms, so it makes less conceptual sense to test a "ground
truth" against these models.

I would not normalize across datasets, and threshold only in terms of
significance (with the appropriate null model), i.e. not threshold in terms
of value.

On Wed, 15 Feb 2023 at 15:39, Danish Mahmood via eeglablist <
eeglablist at sccn.ucsd.edu> wrote:

> Dear Colleagues
> I need to clear few issues in effective connectivity estimation.
> 1. As per my understanding, at one time instance two brain regions (more
> specifically group of neurons) can communicate with each other on a single
> frequency. They may change the frequency but on another time instance. So,
> is it possible to have connectivity connections over multiple frequencies
> for one estimation? (let's say I am taking 2 sec EEG per sample.)
> 2. Is it possible to generate effective connectivity synthetically with
> known strength. I know there are various methods of generating equations to
> produce synthetic EEG data with known connections but I am interested in
> exact strength.
> 3. Is there any robust technique available for normalization and
> thresholding of effectivity connectivity for comparison between different
> datasets?
> Thank you
> RegardsDr Danish M. Khan
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to
> eeglablist-request at sccn.ucsd.edu
>