Initialization Effects in Computer Simulation Experiments.
Abstract
Much progress has been made in dealing with the difficult and important problem of initialization bias. Indeed, simple truncation rules are being replaced by sequential procedures such as those given in Kelton (1980) and Heidelberger and Welch (1983). Various tests to detect the presence of initialization bias are also being developed. The authors feel that more attention should be given to the problem in the decision making and experimental design contexts. Simulation researchers might also study the different issues that are involved in simulation experiments with a single system (or single performance measure) vs. experiments with multiple systems (or multiple performance measures). Finally, simulation initialization errors may have different effects depending on whether the simulations are used for design, optimization, evaluation, selection, or feasibility decisions; some investigation in this area is warranted. Keywords: Applied mathematics; Queueing theory; Confidence intervals.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1984
- Accession Number
- ADA161964
Entities
People
- David Goldsman
- Lee Schruben
Organizations
- Cornell University College of Engineering