Simplifying Statistical Error Modeling of Nuclear Material Measurements

The true amount of measurement error affecting nuclear material accounting measurements is unknown. Statistical models are used to estimate and quantify the measurement uncertainty resulting from random and systematic (bias) sources of measurement error. Specifically identifing and modeling for all measurement error sources affecting an inventory can be incredible, resulting in a large number of very complex statistical model equations. All statistical error models are somewhat debatable given underlying assumptions and the detail needed for adequate approximations. The challenge for an MC&A statistician is to adequately model and quantify measurement error uncertainty without the model becoming too complex, or disregarding the effect of known and unknown error sources. Collectively aggregating the sources of random and systematic error can simplify the statistical error models, making the uncertainty calculations much easier to manage.