Unconstrained sources (EEG) - Chi-squared test results (within-subject) and comparison between groups

Discussions
1

Hello!

I am doing a source localization on EEG data for the difference between two conditions (Rare minus Frequent) - sLORETA, unconstrained sources. The EEG setup was low-density (27 electrodes) as we didn't plan to go beyond EEG, but the data propmted the analysis. I have two questions:

  1. I followed the workflow for the "when and where" question, and as stated in the Tutorial 27, the parametric one-sample Chi-squared-test for unconstrained sources is very sensitive (the whole brain is significant). This is at alpha 0.001, Bonferroni correction. Does it make sense to use the Chi-squared values (e.g., above mean+2 standard deviations, 5% largest or anything else) to sort of define regions with greatest differences or is there a different solution?

  2. I am really uncertain about this one - I also have two groups and I compared the differences (|Rare-Frequent|) between the groups with a non-parametric independent t-test. Is this correct or should I only compare a condition to another condition (e.g., Rare between the groups)?

I am very new to source localization, and I would appreciate any comment!

Kind regards

2

Testing unconstrained source maps is indeed complicated and still not sorted out by the experts in statistics in our team.
@Sylvain @pantazis Could you give some advice at this level?

3

The approaches you describe are fine. As François mentioned, the stats for unconstrained sources are still incomplete in the app. I can recommend you produce constrained source models and follow your proposed approach.

4

Thank you very much, this helps a lot!

1 Like