Using Real-World and Simulation Data to Estimate a Location Parameter

SYNOPTIC ABSTRACT

This paper addresses the problem of estimating a location parameter μ using observations from both a real-world system and from a simulation model of that system in the estimation process. The paper examines, for a simple situation, four possible estimation methods and investigates how bias in the simulation model affects their usefulness. None of the four estimators dominates or is dominated by any other estimator. One, the adaptive estimator appears to be the rational choice since it provides reasonable gains (decreases in MSE) for small biases and, in the worst case, does not result in a substantial penalty.

Key Words and Phrases:

• estimation
• model bias
• combining data