Sharp moderate maximal inequalities for upward skip-free Markov chains (Q2312777)

A continuous-time Markov chain \(X\) on the nonnegative integers is called upward free if it only allows nearest-neighbor birth jumps but allows large death jumps. The paper studies moderate maximal inequalities for upward skip-free Markov chains, which include the \({L^p}\) maximal inequalities for martingales as special cases. In what follows, a function \(F:[0,\infty ) \to [0,\infty) \) is called moderate if (a) it is a continuous increasing function and \(F(0) = 0\); (b) there exists \(\beta > 1\) such that \(\mathop {\sup }\limits_{x > 0} \frac{{F(\beta x)}}{{F(x)}} < \infty \). Let\(X{}_t^* = \mathop {\sup }\limits_{0 \le s \le t} {X_s}\). The main result of this paper gives conditions under which for a function \(g\) defined explicitly in terms of infinitesimal characteristics, and for any moderate function \(F\), there exist two positive constants \({c_F}\) and \({C_F}\) such that for any stopping time \(\tau \) of \(X\), \({c_F}EF(g(\tau ) + 1) \le EF(X_\tau ^* + 1) \le {C_F}EF(g(\tau ) + 1)\). In particular, for any \(p > 0\), there exist two positive constants \({c_p}\) and \({C_p}\) such that for any stopping time \(\tau \) of \(X\), \({c_p}E{(g(\tau ) + 1)^p} \le E{(X_\tau ^* + 1)^p} \le {C_p}E{(g(\tau ) + 1)^p}\). The results are applied to the \(M/M/1\) queue and to the upward skip-free Markov chain with large death jumps.