US and Russian Collaborations on the Analysis of Inventory Differences Using Propagation of Variance

During the past few years there have been several collaborations between US DOE laboratories and facilities in Russia that analyze inventory differences (IDs). A companion paper considers approaches to the historical analysis of IDs. This paper presents approaches to ID analysis using propagation of variance (POV). Our goal for POV is to estimate the measurement-error standard deviation (s ID) of the ID. We focus on one example from the fuel fabrication area at the Electrostal Machine Building facility. The example data was prepared by A. Karpenko and V. Nekrasov. It is part of the ID statistical- evaluation workshop offered at the Russian Methodological Training Center (RMTC) in Obninsk, Russia. In our example, we include the effects of both systematic and random errors. The US facilities normally have access to in-house measurement-control data to estimate random and sys- tematic error variances. Currently, most Russian facilities do not have measurement-control data to provide such estimates. Therefore, Russian facilities typically rely on instrument developers’ claims to separately estimate systematic and random error variances. Once we have reasonable estimates of the main error variances for all instruments, we can perform POV. The tools for performing POV range from calculations made by hand to specialized software such as MAWST. We review the available tools and present short-cut approximate formulas that we expect will be adequate for the Elektrostal example. We give enough statistical detail to allow the reader to understand how assumptions about the measurement processes affect the s ID calculations.