Determining Sample Size in Binary Measurement System
Yan Zhao,Zhen He,Liam Blunt,Xiangqian Jiang,Yanlong Cao,Hongyu Zhang
DOI: https://doi.org/10.1117/12.819773
2008-01-01
Abstract:Measurement system analysis guarantees the reliability of acquired data. Although much research has been performed regarding variable measurement system and Gage R&R has been comprehensively employed across many companies, there is relatively little attention that has been paid to binary measurement system, which is considered to be more practical and efficient. Proportion of agreement is generally utilised to evaluate binary measurement system in the traditional AIAG method. As a consequence, sample size should be more reasonably determined, in which process the number of parts, appraisers, and trials are of key importance. However, this critical issue has not been profoundly investigated as yet. In the present study, the number of parts is determined through the plot of length of confidence interval, and an alternative method is introduced to choose the number of appraisers and trials based on the majority voting rule. This is considered to be more sensible than the prevalent rule, in which two appraisers and two trials are usually chosen and an agreement is made only when the conclusions of both appraisers and trials are the same. In addition, a data set is analysed using the proposed method, and the results indicate that it is more rational.