False Alarm Testing for Radiation Detection Systems
Dennis D. Leber, Leticia S. Pibida
An operator of a radiation detection system that displays a high rate of false alarms may be- come desensitized to these alarms. This action is known as alarm fatigue and can have detri- mental results as the operator may cease to respond to what may be valid alarms. To mitigate these adverse outcomes, an agency may seek to procure radiation detection systems with ad- equately low false alarm rates. In developing a test to confirm that the system satisfies the stated false alarm rate threshold requirement, the number of required observations or test du- ration is often an initial question. In this chapter, we demonstrate how an experimenter can develop a successful false alarm test with two provided pieces of information: the false alarm rate threshold requirement and a statement of acceptable risk or required confidence. Using the statistical hypothesis testing framework, we illustrate the meaning of risk and confidence from both the consumer and producers perspectives and provide guidance on selecting an informed false alarm rate threshold requirement and statement of acceptable risk. We consid-er the binomial and Poisson probability models that apply to testing of radiation detection systems that are employed with and without occupancy sensors, respectively. From these probability models we define the power of a test and demonstrate how an experimenter can use a power curve to balance the tradeoffs between test burden (costs) and producer risk (type II error) while satisfying the required confidence. We provide sample size and ac-ceptance criterion tables to define fixed sample tests that satisfy a variety of false alarm thresholds and levels of acceptable risk for systems with and without occupancy sensors.
and Pibida, L.
False Alarm Testing for Radiation Detection Systems, Technical Note (NIST TN), National Institute of Standards and Technology, Gaithersburg, MD, [online], https://doi.org/10.6028/NIST.TN.2118
(Accessed September 28, 2023)