PERFORMANCE EVALUATION OF SAMPLING PLAN METHODOLOGY

This paper discusses the problem of improving sampling plan methodology for inventory verification. The paper focuses on the relationship between sampling plan methodology and the error behaviour of NDA measurement instruments. The paper points out that optimal sampling plans should be a function of the error behaviour of the instruments, but emphasises that reliable information about this error behaviour is not really available. As a result, existing methodologies are the choice of a robust approach which will give satisfactory detection performance Independent of the instrument error behaviour. The drawback, however, is that in some instrument behaviour scenarios, the ignorance of error behaviour will be paid for by an excessive use of DA measurements which are a principal cost factor. The paper illustrates some of the dilemmas in sampling plan design and traces them to the need for better information about NDA error behaviour. This is used to identify and focus priorities for studies of the performance of NDA instruments.