Analysis of sensitivity of reciprocal matrices (Q798731)

The author is concerned with the relationship between the principal right eigenvectors of a given reciprocal matrix and its permutation by Hadamard (elementwise) multiplication with another reciprocal matrix. Earlier studies have considered additive permutations. A reciprocal matrix, A, is one for whose elements satisfy \(a_{ij}=1/a_{ji}\). A reciprocal matrix is consistent if its elements satisfy the further restriction that \(a_{ij}a_{jk}=a_{ik}\) for each i,j,k. The Perron-Frobenius theorem applies to reciprocal matrices, so they have unique right eigenvectors corresponding to the largest positive real eigenvalue. The author's main result is paraphrased in the following Theorem. Let A denote a consistent reciprocal matrix, P a reciprocal matrix, and \(A\circ P\) their Hadamard product. If a and p are the principal right eigenvectors of A and P respectively, then the principal right eigenvector of \(A\circ P\) is \(a\circ p\). If A and P are 3\(\times 3\) matrices, then the same result holds for arbitrary reciprocal A. In order to show this result, the author first demonstrates the following Theorem. Let A be a reciprocal matrix. A can be uniquely decomposed as \(A=H\circ E\), where H is a consistent reciprocal matrix with the same principal right eigenvector, and E is a reciprocal matrix with the same largest eigenvalue. These results are illustrated by examples and extended to sequences of perturbations.