Estimation of Traffic Intensity Vector of a Two Stage Open Queueing Network Models with Feedback

Abstract

We propose a consistent and asymptotically normal estimator for traffic intensity vector P of a two stage open queueing network model with feedback with no assumption of arrival and service time distribution. Using this estimator and its estimated covariance matrix A, a 100(1- a )% confidence region for traffic intensity vector is constructed. Standard bootstrap, Bayesian bootstrap and percentile bootstrap are applied to develop the confidence regions. Simulation study was undertaken to evaluate performance of the four confidence regions. The performances are assessed in terms of their coverage area percentage, average area and relative coverage area. Further calibration technique is used to improve the coverage area percentages of confidence regions.