Abstract
In simulation an input variable like interarrival time is sampled, and hence its average deviates from its known expectation. This information can be used to improve the estimated simulation response: regression sampling or control variate technique. The usual crude estimator is shown to be biased. If local linearity holds, then the regression estimator becomes unbiased. Moreover its variance becomes smaller under mild conditions. The assumption of local linearity is an alternative to the normality assumption of other authors. This paper further emphasizes the difference between results ex ante (unconditional) and ex post (given the experimental input values). A telephone-exchange simulation provides a case study.
Similar content being viewed by others
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Hopmans, A., Kleijnen, J. Regression Estimation in Simulation. J Oper Res Soc 31, 1033–1038 (1980). https://doi.org/10.1057/jors.1980.190
Published:
Issue Date:
DOI: https://doi.org/10.1057/jors.1980.190