Thank you Tim and Saad for your help!
I think I found a nice measure to check the effect on the distributions on
phi and theta. I look at your code of make_dyadic_vectors and I liked the
dyadic tensor approach. I use now the eigenvalues of the mean dyadic
tensor (using phi and theta) for each voxel to compute a dispersion or
coherence value.
Cheers,
Jeroen
|