Testing for Nonlinear Dependence in Inventory Data

Abstract

This paper describes the properties of nonlinear dynamical systems in demand data and assesses the implications for inventory management. A test from Brock, Dechert, and Scheinkman (BDS) [6] for detecting nonlinearities is presented. This test assists inventory managers with two issues. First, it can be used to detect departures from the independence and stationarity assumptions of particular inventory models, thereby identifying the need for more suitable models. Second, the BDS test can determine whether forecasting model improvements are possible, and can measure improvements in the forecasting model process. The corresponding reductions in forecast errors improve inventory management and lead to reduced inventory costs. Using actual weekly demand for oil filters, this paper demonstrates the use of the BDS test and the effects of nonlinearities in demand data on inventory model performance.