A New Approach for Stratified Sampling in Inventory Cost Estimation.

Model-based statistical sampling (MBSS) is a recently developed method for designing efficiently stratified sampling plans to estimate a total. MBSS has been suggested by the authors for inventory cost estimation. In a previous paper, the mathematical formulation of MBSS was given in detail and one example was presented to demonstrate the feasibility of the approach. In the present paper, MBSS is discussed with greater attention given to issues relevant to auditors. Four inventory populations are analyzed and sampling plans are designed to meet auditing standards. Comparisons are made between model-based stratification, Dalenius-Hodges stratification, equal-aggregate-size stratification, and proportional allocation. On average, the model-based sampling plans require 12 percent fewer items than the Dalenius-Hodges sampling plans for the same planned levels of precision and reliability. Simulation results confirm the achieved precision and reliability of these procedures but reveal sometimes poor coverage for confidence intervals, even in a population with a high error rate.