Computer simulation of the Sequential Probability Ratio Test for nuclear safeguards

A Fortran IV computer program called SPRTEST is used to simulate the Sequential Probability Ratio Test (SPRT). The program provides considerably more information than one can obtain from the approximate SPRT theory of Wald. For nuclear safeguards applications SPRTEST permits the equipment designer to optimize the input test parameters and, indeed, to determine whether the SPRT is the statistical test of choice. Using Monte Carlo techniques, SPRTEST simulates the use of the SPRT in a radiation monitor. The accumulation of monitoring data from a normal distribution is simulated by repeated sampling of a random number generator. In this way, SPRTEST determines the expected false-positive (α) and false-negative (β) detection probabilities and the average step number (ASN) for a particular SPRT. The report describes SPRTEST, provides a Fortran listing, and demonstrates SPRTEST applications. The report also compares results with those expected from the single-interval test (SIT) on which the SPRT is based; generally, the SPRT provides better detection probabilities for a wide range of source strengths and, at bakcground levels, it takes less time, on average, to make decisions. To obtain optimal results with the SPRT, it must have the capability to detain the counting subject for longer than the SIT time. The SPRTEST program should be useful in choosing the best statistical test for a wide variety of applications, including safeguards, health physics monitoring, and general nuclear detection. 14 refs., 11 figs., 15 tabs