The model is coded in SLAM II with the use of extensive FORTRAN subroutines.
In order to avoid bias from the transient stage, Welch's graphic procedure (Law and Kelton, 1992) was used to determine when the transient phase ended.
The results showed that the transient phase ended after approximately 1560 h of simulated time and to be conservative that time was then doubled.
The statistical arrays were cleared at the end of the transient phase and the model was allowed to run for an additional 18720h. Running the model for this amount of time allowed enough observations to be collected so that the data sets met the requirements of normality.
The model used synchronized random number streams and common random numbers as variance reduction techniques (Kleijnen, 1987).
Complete details will be presented in the experimental design section.