Heuristic Rules for Improving Quality of Results from Sequential Stochastic Discrete-Event Simulation (2012)
AuthorsPawlikowski, K., McNickle, D., Lee, J.-S. R.show all
Sequential analysis of output data during stochastic discrete-event simulation is a very effective practical way of controlling statistical errors of final simulation results. Such stochastic sequential simulation evolves along a sequence of consecutive checkpoints at which the accuracy of estimates, usually conveniently measured by the relative statistical error, defined as the ratio of the half-width of a given confidence interval (at an assumed confidence level) to the point estimate, is assessed. The simulation is stopped when the error reaches a satisfactorily low value. One of problems with this simulation scenario is that the inherently random nature of the output data produced during a stochastic simulation can lead to accidental, temporary satisfaction of the stopping rule. Such premature stoppings of simulations is one of causes of inaccurate final results, producing biased point estimates, with confidence intervals that do not contain the exact theoretical values. In this paper we consider a number of rules of thumb that can enhance the quality of the results from sequential stochastic simulation despite that some simulations can be prematurely stopped. The effectiveness of these rules of thumb is quantitatively assessed on the basis of experimental results obtained from fully automated simulations aimed at estimation of steady-state mean values.
Keywordscoverage of confidence intervals; sequential stopping rules; statistical errors of results; stochastic discrete-event simulation
ANZSRC Fields of Research08 - Information and Computing Sciences::0801 - Artificial Intelligence and Image Processing::080110 - Simulation and Modelling
01 - Mathematical Sciences::0104 - Statistics::010406 - Stochastic Analysis and Modelling
01 - Mathematical Sciences::0104 - Statistics::010401 - Applied Statistics