DOE treats the simulation model as a black box; i.e., only the inputs and outputs are observed and analyzed. For example, in the simulation of the statistic (in Sect. 3.2) the simulation inputs (listed in Step 1) are (mean), (variance), (sample size), and (number of macro-replicates); this is probably a tactical factor that is not of interest to the user. Suppose the user is interested in the quantile of the distribution function of the statistic in case of nonnormality. A black box representation of this example is:
One possible metamodel of the black box model in (3.3) is a Taylor series approximation - cut off after the first-order effects of the three factors, :
Besides the metamodel specified in (3.4), there are many alternative metamodels. For example, taking the logarithm of the inputs and outputs in (3.4) makes the first-order polynomial approximate relative changes; i.e., the parameters , , and become elasticity coefficients.
There are many - more complex - types of metamodels. Examples are Kriging models, neural nets, radial basis functions, splines, support vector regression, and wavelets; see the various chapters in Part III - especially Chaps. III.5 (by Loader), III.7 (Müller), III.8 (Cizek), and III.15 (Laskov and Müller) - and also Clarke, Griebsch, and Simpson (2003) and Antioniadis and Pham (1998). I, however, will focus on two types that have established a track record in simulation:
To estimate the parameters of whatever metamodel, the analysts must experiment with the simulation model; i.e., they must change the inputs (or factors) of the simulation, run the simulation, and analyze the resulting input/output data. This experimentation is the topic of the next sections.