Estimating Shipper/Receiver Measurement Error Variances by Use of ANOVA

Every measurement made on nuclear material items is subject to measurement errors which are inherent variations in the measurement process that cause the measured value to differ from the "true " value. In practice, it is important to know the variance (or standard deviation) in these measurement errors, because this indicates the precision in reported results. If a nuclear material facility is generating "paired" data (e.g., shipper/receiver) where party 1 and party 2 each make independent measurements on the same items, the measurement error variance associated with both parties can be extracted. This paper presents a straightforward method for the use of standard statistical computer packages, with analysis of variance (ANOVA), to obtain valid estimates of measurement variances. Also, with the help of the P-value, significant biases between the two parties can be directly detected without reference to an F-table.