Realizable monotonicity for continuous-time Markov processes (Q981007)

The authors analyze the notions of \textit{stochastic monotonicity} and \textit{realizable monotonicity} for continuous time Markov chains with a state space \(S\) being a finite partially ordered set. A Markov chain \(\eta_t\) on \(S\) with transition probabilities \(P_t(x,y)=P(\eta_t=y\mid\eta_0=x)\) is said to be \textit{stochastically monotone} if for each pair \(w,z\in S\) with \(w\leq z\) there exists a Markov chain \(X_t(w,z)\) on \(S\times S\) such that (i) \(X_0(w,z)=(w,z)\); (ii) each component \(X_t^i(w,z),\) \(i=1,2,\) is a Markov chain on \(S\) with transition probabilities \(P_t\); (iii) \(X_t^1(w,z)\leq X_t^2(w,z)\) for all \(t\geq 0.\) A Markov chain \(\eta_t\) on \(S\) with transition probabilities \(P_t(x,y)=P(\eta_t=y\mid\eta_0=x)\) is said to be \textit{realizably monotone} if there exists a Markov chain \(\xi_t\) on \(S^S\) such that (i) \(\xi_0=\mathrm{Id}\); (ii) for every \(z\in S,\) the process \(\xi_t(z)\) is a Markov chain with transition probabilities \(P_t\); (iii) if \(w\leq z\) then \(\xi_t(w)\leq\xi_t(z)\) for every \(t\geq 0.\) Clearly, realizable monotonicity implies stochastic monotonicity. For a discrete time Markov chain the partially ordered finite sets \(S\) for which the converse is true were characterized in [\textit{J. A. Fill} and \textit{M. Machida}, Ann. Probab. 29, No. 2, 938--978 (2001; Zbl 1015.60010)] in terms of \textit{Hasse diagrams} (\textit{cover graphs}) of \(S.\) For continuous time Markov chains possessing an infinitesimal generator the authors prove that if in a poset \(S\) stochastic monotonicity implies realizable monotonicity in discrete time, then the same holds true in continuous time case. They show that the converse is not true. Using computer-assisted method the authors find a complete list of five and six-point posets for which the above equivalence fails. Moreover, they prove that in each poset containing one from the list as an \textit{induced sub-poset}, equivalence fails.