10.2 Parameter Estimation

To run a simulation program, the values of certain input parameters are needed, such as failure and repair rates. In addition, we need a way of analyzing the simulation output and extracting parameters such as reliability and mean time to system failure. In this section we will see how such parameter values can be estimated. We will distinguish between point estimation and interval estimation, describe three methods by which to obtain point estimates of parameter values, and show how a confidence interval for the parameter can be constructed. Most of our discussion assumes that we know the underlying shape of the distribution that the data will follow and that this shape depends on one or more parameters whose exact value is unknown to us. For example, we may believe that processors fail according to a Poisson process, which we can characterize by estimating the rate, ?, of this process. In some cases, we will estimate parameters even without knowledge of the exact shape of the distribution, using approximating formulas (most notably, the Central Limit Theorem).