A Test of a Statistical Method for Computing Selected Inventory Model Characteristics by Simulation

This article is a companion piece to M. A. Geisler, The Sizes of Simulation Samples Required to Compute Certain Inventory Characteristics with Stated Precision and Confidence, MANAGEMENT SCIENCE, January, 1964. Special statistical methods were used in that article to compute the sample sizes for specified inventory models. In this article, the methods were tested by applying them to particular inventory cases, and determining how well the actual precision and confidence obtained in the estimates agreed with expectation. We found that the actual precision and confidence obtained for each of the inventory policies and procurement lead-time cases tested did correspond closely with expectation. Thus, it was expected that 95 per cent of the samples of the zero procurement lead-time cases tried would have sample values within 100 per cent of the true mean value. In the shortage calculations, the results ranged from 94.3 to 97.6 per cent, and for the overages, the range was from 95.6 to 99.9 per cent. From this test, we concluded that the statistical procedure given in the prior article is valid, and produces reliable estimates of mean shortages and overages per period.