Convergence issues in the theory and practice of iterative aggregation/disaggregation methods (Q964104)

Iterative aggregation/disaggregation (IAD) methods for the computation of stationary probability vectors of large scale Markov chains form efficient practical analysis tools. In this paper, some new concepts are studied as well as their behavior and dependence on the parameters involved in aggregation algorithms, and establish some necessary and/or sufficient conditions for convergence. The theory developed offers a proof of convergence of IAD algorithms independent of whether the governing iteration matrix is primitive or cyclic of two IAD processes.