Some Asymptotic Formulas for Markov Chain with Applications to Simulation
P. W. Glynn
Journal of Statistical Computation and Simulation, Vol. 19, 97-112 (1984)
Various techniques have been proposed for determination of confidence intervals associated with steady-state quantities in simulation. Evaluation of such procedures requires comparison of their performance on stochastic systems with known characteristics. In this paper, the authors therefore derive computable formulas for the initial bias, variance and spectrum of the sample mean in finite state Markov chains, and discuss their relevance to the steady-state simulation problem.