First, define and under this rotation, the dynamics are given as where Since and are similar, they share the same eigenvalues. And since we are assuming that has the real eigenvalues within the unit circle, also does.
Now consider the Schur decomposition (see Schur's lemmaSchur's lemma) of where is a complex Hermitian matrix and is an upper triangular matrix with diagonals equal to the eigenvalues of . As the eigenvalues of has real-valued eigenvalues, and are both real-valued (implying that ).
Also, by implication, is an orthogonal matrix (i.e. ), and we have the equality of where is a lower triangular matrix.
Define the rotation and similarly, the dynamics become where since due to the orthogonality of . This means that the variance remains unchanged under compared to , implying that Now finally, consider the translation where is well-defined since none of the eigenvalues of are equal to . Under , the dynamics become where Thus we have where we have , , and is real-valued lower triangular matrix. Lastly to make every elements of be non-negative, we can rotate the factors one last time so that if is negative, then is multiplied by . Therefore the resulting model satisfies the restrictions of Identification of ATSMs > ^026429Identification of ATSMs > Definition 2 (Dai-Singleton model).