Multivariate Batch Means and Control Variates

We consider applying the nonoverlapping batch means output analysis method in conjunction with the control-variate variance-reduction technique to estimate a steady-state multivariate mean vector. The effects of the number of batches and the number of control variates on the multivariate point and region estimators and the univariate point and interval estimators are considered. The results are experiment analysis guidelines in terms of an appropriate range of the number of batches to choose as a function of the number of responses and control variates. The results have implications for terminating simulations as well.

• 저널Management Science · 38(10) · 1415–1431
• 토픽Control Systems and Identification · Control and Systems Engineering
• DOI10.1287/mnsc.38.10.1415
• 저자Wei‐Ning Yang, Barry L. Nelson