Calculation of Inventory Uncertainty by Estimate Sampling

In answer to the question, \"How accurate is the physical inventory data just compiled?\", a method has been developed which states in grams (or other physical units) the most probable value for the true inventory and the statistical confidence interval within which the true inventory probably lies. Examples drawn from use of the method are discussed. The method is compared to traditional methods such as Gaussian propagation of error, and either attribute or variables inspection for acceptance, in order to show that each method basically answers a different question. Choice of any given method should be made primarily by consideration of the exact question to be answered. Generally referred to as \"estimate sampling,\" the recommended method determines the average difference between the data used on the inventory and the \"true\" data (defined as the data resulting from a careful redetermination of selected items, using the best methods available,) Stratification of the inventory is not necessary but general y is desirable. Sample sizes are not critical, and typically are no mora than half the number of samples needed for good attribute sampling.