|
Abstract:
|
Proportion of nonconformance is the basis for calculating the percentage of parts or the parts per million (ppm) outside the specification limits. These key metrics are frequently used to assess the quality of a process. Because the proportion is almost always computed using a random sample from a larger population, the resulted proportion value is a point estimate that is subject to random sampling error. Therefore, to properly assess the true proportion of nonconformance for all the items from a process, a reliable method is needed to estimate the upper confidence limit for the proportion of nonconformance with a high confidence level. Our study proposes a new formula using the delta method and the distribution properties of the sample mean and the sample variance from a normal process. By extensive simulations, we compare the accuracy and precision of the new method with three existing ones, and report the improved performance of the proposed method for various settings. Based on our simulation results, we recommend a hybrid approach from the proposed method and an existing method to construct the upper confidence limit for all cases.
|
