Simple Acceptance Decisions and Acceptance Risk

Abstract

Acceptance risk may be estimated using cumulative probability distribution functions applied to populations of measurands and test measurement values. “Simple acceptance” is a decision rule that sets the acceptance range of a test result equal to the tolerance range specification. While acceptance risk is comprised of both consumer risk and producer risk, this paper compares the effects of simple acceptance decision rules and guardbanding decision rules on consumer risk. Consumer risk is also known as the probability of false acceptance. The terms describe the risk of accepting test results as passing when the actual values exceed specification limits. False acceptance is only possible when the true value of a measurand is out of tolerance and the test result indicates that the measurand is within tolerance. Metrologists generally have some information about the test uncertainty regarding a specific acceptance test result. Along with a general lack of knowledge regarding the measurand population parameters, the complicated interplay between risk, dispersion, and central values generally prevents calibration laboratories from fully characterizing acceptance probabilities. Organizations that model the measurand populations can reduce consumer risk by avoiding certification of items whose measurand populations are not well centered. The models in this paper present a recurring trend: guardbanding reduces nonnegligible risks of false acceptance when compared to simple acceptance. However, guardbanding is not the most effective means of acceptance risk mitigation. Systematic characterization of measurand populations can provide the information a calibration laboratory needs to reliably control acceptance risk.