<t>I agree with Reza's point on positive definiteness. IfX = (X_i), v_i = Var(X_i), c_ij = Cov(X_i, Y_j), C = (c_ij)Then the correlation matrix Corr(X) = V C V, where V = diag{1/sqrt(vi)}. So if the covariance matrix is pd (or psd), so is the correlation matrix. Can test for positive definiteness by...