[Eeglablist] Statistics on Global ERSP Average of Independent Components in Single Cluster

[Zaeem Hadi]

Dear EEGLAB experts,
I have a 32-channel eeg dataset with 15 participants and I am interested in within-subject ERSP differences in two experimental conditions.
For the STUDY, I only compute a single cluster and then manually reassign the noisy independent components (ICs) to a manually created "noise" cluster.
I have about 150 ICs in the relevant single cluster after separating the noise ICs. This contains components from all the subjects.

My question is if I perform statistics on this single cluster of 150 ICs, to determine the difference between two experimental conditions, what would this cluster-level comparison represent? Could this be called a global ERSP representation?Importantly in this situation, every single independent component would be considered an independent sample during statistics and probably it artificially inflates statistical power. I was wondering that what is the best way to deal with this and if this is a sound way to proceed with statistics.

Note that I do not want to create more clusters as I am only interested in a global ERSP representation. Even with 2 or 3 clusters, there would be sufficient overlap such that each will contain multiple ICs from the same subject. How does one deal with the artificially inflated statistical power when performing IC level statistical analysis?


I would greatly appreciate your help on this.
Thanks also for responding to my previous queries.
Best wishes,
Zaeem