Title: Steady State Sensitivity Analysis of Continuous Time Markov Chains

Abstract: In this paper, we study Monte Carlo estimators based on the likelihood ratio (LR) approach for steady-state sensitivity. We extend the result of Glynn and Olvera-Cravioto [arXiv:1707.02659] to the setting of continuous time Markov chains (CTMC) with a countable state space which include models such as stochastic reaction kinetics and queuing networks. We prove that the LR approach is asymptotically unbiased. Furthermore, we show that the variance of the centered LR (CLR) estimators does not grow in time. This result suggests that the CLR estimators should be favored when the mixing time of the CTMC is large. We demonstrate a practical implication of this analysis on a numerical benchmark of a stochastic reaction network.