Hi Jules,

the design matrix appears to be correct. Regarding the connectivity measurse, I would think that the continuos variant of the measure may be more appropriate and better matched to the assumptions of the statistical test. While the binary variant would still work, it might not be as sensitive to detecting changes. 

Andrew

Originally posted by Jules Mitchell:

Hi Dr Zalesky, 

I'm utilising the dNBS toolbox to investigate group (responders vs. non-responders) by time (pre and post-treatment) interactions using an F-test. Thank you for making this extension of NBS available!

I've set-up my design matrix and contrasts as per a previous post you said was okay (https://www.nitrc.org/forum/forum.php?th...), however, I wanted to verify I adjusted this for my purposes correctly. 

I realise I need to drop the group column, and for group comparisons I will perform a t-test between groups separately for each time point.

On a related note, my directed network (32x32 eeg channels) was calculated using multivariate transfer entropy toolbox. The edges in the matrix can contain either a binary [0,1] for no significant information transfer or significant transfer, or it can contain the information transfer value (e.g. 0.1). Is assessing either of these fine or is one more appropriate/powerful? 

Kind regards,

Jules