[Eeglablist] Phase locking value - statistical analysis

[Jung Hwa Han]

Hello all,

I am working with EEG data collected from normal / ultra-high risk for
psychosis / schizophrenia groups during working memory task (3 task loads).

I computed the phase locking values (PLV: Lachaux, 1999) for all channel
pairs (60 channels --> 1770 pairs), and used nonparametric statistics to
look at the differences between groups, and differences between task loads.
I am correcting for multiple (1770) comparisons using FDR.

An issue I am facing is that there seems to be no way to look at the
interaction of the two factors (group & task load) when working with
nonparametric statistics. My questions are below:

1. Does anyone know of any nonparametric method to look at interactions of
2+ factors?
2. Is there a reason to think that applying Fisher-z transform to
phase-locking values may be problematic? I'd like to try this, so that I
can just use parametric tests (2-way mixed model ANOVA). I've seen several
studies that dealt with "coherence" where they used Fisher-z transforms to
normalize (?) the coherence values, then used parametric tests for
statistical analysis.

It's been a painfully long way learning EEG and analyzing and coming this
far! I would appreciate ANY help or any related tips from anyone. Thanks!

Best,

Jung
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20171219/6eedc5ca/attachment.html>