Stochastic Nature of Output Data

• Simualtion result is generated for a given input data. Essentially a model is an input/output transformation.
• Since the input variables are random variables, the output variables are random variables as well, thus they are stochastic (probabilistic).
• Example 12.1 (Able and Baker, revisited)
  • Instead of a single run of simulation, four runs were conducted. The results are shown in Table 12.1 on page 431.
  • The utilization $\hat{\rho}_r$ and system time $\hat{w}_r$ were listed as 0.808,0.875,0.708, 0.842, and 3.74, 4.53, 3.84, 3.98.
  • There are two general questions we have to address by a statistical analysis of the observed utilization $\hat{\rho}_r$
    1. Estimation of the true utilization $\rho = E(\hat{\rho}_r)$ by a single value, called a point estimate.
    2. Estimation of the error in our point estimate, either in the form of a standard error or confidence interval. This is called an interval estimate.
• Example 12.2 (Effect of correlation and initial condition) We will see actual measurement of performance in Section 12.3 and 12.4.
• Example 12.3 (Fifth National Bank of Jasper, revised)