Current Issues in Frequency Domain Experimentation

This paper presents an examination of certain issues in Frequency Domain Experimentation (FDE) for discrete event simulation. Experimental results are presented to demonstrate that (i) conclusions drawn from FDE are dependent on the oscillation frequency and unless frequencies are chosen carefully, misleading results can be obtained; (ii) interpretation of results from frequency domain experiments are fundamentally different from the interpretation of results from regression analysis---specifically, a term found significant by FDE is not to be interpreted as a term able to explain a significant portion of the variation in the response over the experimental region; and (iii) basic assumptions required for FDE, in particular the assumption that input and output processes constitute a time-invariant linear system, do not hold for M/M/1 queues. Other issues discussed are indexing, run length determination, computational effort, and why FDE is not applicable for terminating simulations. It is concluded that further developments are needed before FDE can be used by practitioners.