A Multiple Ranking Procedure Adapted to Discrete-Event Simulation.
AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OH
Pagination or Media Count:
The main objective of this research is to extend the multiple ranking procedure of Dudewicz and Dalal to the case of K normal covariance-stationary processes with unknown and nonidentical covariance structures. To implement this procedure, the author developed a computer program that can be called on the fly by an ongoing discrete-event simulation in order to select the best steady-state performance among K alternative policies. For each alternative policy, the proposed support package analyzes the simulation-generated output series to determine if the accumulated sample size is sufficient to meet the predetermined probability requirement that is, the probability of yeilding the correct selection. If the run length is not sufficient, the program determines that number of additional observations to be taken. Upon reaching an acceptable run length, the program reports the estimated mean response of the policy currently under consideration. The run for each alternative policy is executed independently and does not require any information from the runs for other policies. Author
- Statistics and Probability
- Computer Programming and Software