PROPAGATION OF NON-CONSTANT BIAS CORRECTIONS IN NUCLEAR MATERIALS ACCOUNTING

This paper considers measurement method with linear (or other non-constant) bias. The random errors associated with the method are assumed to have finite range. With this error model the paper considers the correction of the measurement by a previously established calibration curve and the propagation of the systematic error of the corrected observation by means of Taylor series approximation for nuclear materials accountancy or auditing purposes. It is shown that the validity of the error propagation technique depends on the precision of the experiment used to estimate the calibration curve. The paper provides practical criteria for the design of experiments to estimate calibration curves, useful for material control.