Granger causality implementation


the traditional way of implementing GC (i.e. comparing a full and a reduced model), leads to some issues, pointed out to in this paper Stokes, P. A., & Purdon, P. L. (2017). A study of problems encountered in Granger causality analysis from a neuroscience perspective. Proceedings of the national academy of sciences , 114 (34), E7063-E7072.

The main issue is that the reduced model is VAR, the full model is VARMA; the full model has a given model order, the reduced model has order infinite by definition.
Luckily for the field, these issues had been already addressed prior to the Stokes&Purdon paper. One efficient way to solve them is by means of state space models.
More info (and pointers to code) here

I would be happy to help, I am not very familiar with the Brainstorm way of coding and dealing with data.


it sounds promissing!!!
I think we should use it

Thanks very much for your post: we are looking into this and will get back to everyone ASAP.