Reliable results from stochastic simulation models

Abstract

Development of a computer simulation model is usually done without fully considering how long the model should run (e.g. computer time) before the results are reliable. However construction of confidence intervals (CI) about critical output parameters from the simulation model makes it possible to determine the point where model results are reliable. If the results are not reliable, due to the variability within the model, the model user at least knows the degree of confidence that may be placed in the results. Such factors as: 1) model variability, 2) prescribed width of the CI, and 3) the level of confidence of the simulation model results, have a profound effect on the amount of computer time needed to obtain reliable results. Since proper choice of these parameters can significantly affect the time and cost needed to obtain the results from the model, it is imperative that the user of the model consider these factors.

Citation

Gochenour, Donald L., Jr.; Johnson, Leonard R. 1973. Reliable results from stochastic simulation models. Res. Pap. NE-277. Upper Darby, PA: U.S. Department of Agriculture, Forest Service, Northeastern Forest Experiment Station. 5p.