A comparison of selective initialization bias elimination methods

When simulating a non-terminating system, the issue of initialization bias must be addressed. Many approaches have been developed to remove initialization bias from the output data. This paper provides a comparison of 5 selected methods applied to two slightly different 2-machine flow shop models. The experiment tests for statistical differences between mean and variance of the data used by each method to calculate steady state performance measures. Additionally, for each method, the practicality and ease-of-use for general applicability in larger modeled environments is discussed.