Dear Dr. Nieto-Castanon

I was trying to replicate the wRRC method described in your book (p.38) by using LSCOV function for WLS regression.

Suppose, we have two time-series ROI1 and ROI2, and the temporal weighting function W.

I tried to run this code:

ROI1 = [1 3 1 23 4 2 0 -2 -20 4 2 9 21]';
ROI2 = [2 3 0 15 2 0 -5 -3 -10 2 0 5 -5]';
W = [1 1 0.5 -0.1 0 1 1 0 -1 1 1 1 0]';

% WLS
% X = lscov(A,B,W) :: X minimizes (B - A*X)'*diag(W)*(B - A*X)
beta_12 = lscov(ROI2,ROI1,abs(W));
beta_21 = lscov(ROI1,ROI2,abs(W));

Z_12 = atanh(beta_12*sqrt(var(ROI1)/var(ROI2)));
Z_21 = atanh(beta_21*sqrt(var(ROI2)/var(ROI1)));

First, how do you avoid complex numbers afrer Fisher's Z transformation?

Second, shouldn't this produce asymmetric matrices?

Thank you in advance!