Lumping in Markov set-chains (Q1327552)

Let \(S_ 0\) be a convex polytope of \(1\times n\) stochastic vectors. The Markov set-chain is defined as a sequence \((S_ n)_{n\geq 0}\) with \[ S_ n=\{xA_ 1A_ 2\ldots A_ n: x\in S_ 0,\;A_ i\in I\},\qquad n\geq 1, \] where \(I\) is a set of \(n\times n\) stochastic matrices \(A\) such that \(P\leq A\leq Q\) for fixed nonnegative matrices \(P\) and \(Q\). The main result of the paper states that the compatible lumping of the matrices \(P\) and \(Q\) induces the lumping of the Markov set-chain (i.e. the corresponding lumping of \((S_ n)\) generates a new Markov set-chain). A certain application of the result is also considered.