Statistical Estimation Problems in Inventory Control

In this paper we define and illustrate the use of the concept of "Expected Total Operating Cost" (ETOC) in dealing with inventory policy estimation when the cost structure is piecewise linear. We show, through examples, that estimates based on classical procedures are often unsatisfactory when viewed in terms of ETOC, and derive improved estimates. Then we look at the problem from a Bayesian point of view and derive the prior distributions (within a particular class) that are implied by the adoption of these superior procedures.