[FieldTrip] Granger Causality statistics and group analysis

[Jörn M. Horschig]

Dear Helen,

 

Great to hear that you liked the paper! 

 

When computing Granger causality you get by default two values, x->y and y->x. I am not sure what more to say to answer your question, so here’s a slightly more detailed response on what we did to the data ;)

 

We did statistics in a two step procedure. First, identifying regions of interest using a surrogate distribution and a permutation test on the z-transformed coherence metric. Second, computing the Granger causality metric and performing an ANOVA (after log-transforming the Granger values to make them satisfy the normality requirement). For both steps, I used FieldTrip functions only for computing the connectivity metric (ordinary coherence, and Granger causality, respectively). Everything else is something I am afraid you will have to do manually (but you can grab some bits and pieces of existing FT functions). I do not have access to my old drive at the Donders Institute anymore, so I cannot send you the code, but I am confident that you can manage to re-code this :) If you get stuck, however, feel free to let me know and I can give you a hand.

 

With best regards,

Jörn

--
Jörn M. Horschig, PhD, Software Engineer
 <http://www.artinis.com/> Artinis Medical Systems  |  +31 481 350 980 

 

From: Helen Wieffering [mailto:helen.wieffering at gmail.com] 
Sent: Monday, February 29, 2016 1:26
To: FieldTrip discussion list <fieldtrip at science.ru.nl>; jorn at artinis.com
Subject: Granger Causality statistics and group analysis

 

Dear Jörn, 

I recently re-read your 2015 paper on Directed Communication between Nucleus Accumbens and Neocortex in Humans. I am taking a similar approach with my data, looking at non-parametric Granger Causality as computed through FieldTrip. I believe you pointed me to this paper in a response to one of my FieldTrip e-mails a few months back, and it has proven very helpful!

If it's not too much trouble, I now wonder if you could provide any further detail on how you computed p-value statistics from each subject's Granger spectrum. As far as I can tell, Fieldtrip's statistical packages are very limited / nonexistent when it comes to connectivity data. In your paper you mentioned obtaining one Granger estimate for each direction - could you explain how you did this, and whether you performed this statistical analysis in Fieldtrip? 

Any advice is much appreciated! Thank you,

Helen 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20160229/dc0a4f23/attachment-0002.html>