Abstract
under which stochastic service times of a simulation model could be replaced by the mean of the probability distribution. This problem is solved by identifying the conditions and the various factors which influence the simulation model. The study was conducted at various levels of the identified factors to analyze the sensitivity of the model. This experiment is done for terminating and non-terminating systems. Two quantitative and three qualitative factors each at different levels are identified. A factorial experiment approach is used since they are useful for examining the effects of two or more factors on a response variable, whether or not interaction exists. The objective is to compare the point estimate of the measured performance, namely time in system, of the entities from the model having constant service time with that from the stochastic model. This is accomplished by developing a confidence interval for the difference in the measured performance between standard model and stochastic model, which is referred as the paired-t test. The approach chosen guides an analyst by specifying the conditions and constraints under which the distribution of service times in a stochastic model could be replaced by the mean of the distribution, helping the analyst to avoid the process of fitting probability distributions from raw data.
Balasubramanian, Udayakumar (1993). A study of deterministic vs. stochastic service times using simulation. Master's thesis, Texas A&M University. Available electronically from
https : / /hdl .handle .net /1969 .1 /ETD -TAMU -1993 -THESIS -B171.