Assessing a binary measurement system with an underlying continuous measurand using targeted verification

Abstract Suppose an important continuous quality characteristic with specification is expensive to measure with a gold standard measurement system. A 100% pass/fail inspection scheme uses a binary measurement system such as a no-go gauge to avoid the expensive gold standard measurements. The inspection scheme makes some errors and we are interested in estimating both the probability of passing a bad part and of failing a good part. We assume that the inspection system is not destructive so we can inspect parts multiple times if we so choose. A part is verified if we use the gold standard system to determine if the part is within specification or not. We propose and quantify the benefits of a new cost-effective assessment plan that verifies only a small fraction of the parts selected for the study.