Discrete Time Methods for Simulating Continuous Time Markov Chains.
STANFORD UNIV CALIF DEPT OF OPERATIONS RESEARCH
Pagination or Media Count:
The regenerative method for analyzing simulations of positive recurrent, continuous time Markov chains yields confidence intervals for parameters associated with the stationary distribution of the Markov chain. In this paper two methods are developed which permit the simulator to simulate discrete time Markov chains and still produce confidence intervals for the original continuous time Markov chain. These methods are computationally more efficient in that exponential holding times need not be generated. Furthermore, one of the methods is also statistically more efficient in that it leads to a smaller variance constant in the resulting confidence interval.
- Statistics and Probability