Variance Reduction Techniques in the Simulation of Markov Processes
Oliviero Lessi · Joint Research Centre (European Commission) · 1987
WE STUDY A FUNCTIONAL R OF THE STATIONARY DISTRIBUTION OF A HOMOGENEOUS MARKOV CHAIN. IT IS OFTEN DIFFICULT OR IMPOSSIBLE TO PERFORM THE ANALYTICAL CALCULATION OF R AND SO IT IS REASONABLE TO ESTIMATE R BY A SIMULATION PROCESS. A CONSISTENT ESTIMATOR R(N) OF R IS OBTAINED WITH RESPECT TO A CHAIN WITH A COUNTABLE STATE SPACE. SUITABLY MODIFYING THE ESTIMATOR R(N) OF R ONE OBTAINS A NEW CONSISTENT ESTIMATOR WHICH HAS A SMALLER VARIANCE THAN R(N). THE SAME IS OBTAINED IN THE CASE OF FINITE STATE SPACE.