Transition matrices for well-conditioned Markov chains (Q884415): Difference between revisions

The authors study the class of irreducible stochastic matrices \(T\) of order \(n\) such that \(\kappa_3(T)=\frac {n-1}{2n}\), where \(\kappa_3(T):=\frac 12\max_{1\leq j\leq n}\pi_j\| (I-T)_j^{-1}\| _{\infty}\), \((I-T)_j\), \(j=1,2,\dots, n\) the pricipal submatrices of \(I-T\) by deleting the \(j\)th row and \(j\)th column of \(I-T\), \(I\) the identity matrix and \(\pi\) the stationary distribution vector of the finite ergodic homogeneous Markov chain defined by \(T\) (as the probability transition matrix). Such matrices correspond to Markov chains with desirable conditioning properties. They identify some restrictions on the zero-nonzero patterns of such matrices, and construct several infinite classes of matrices for which \(\kappa_3\) is as small as possible.