Robust Quality Analysis Using Coarsely Discretized Measurements

摘要

In many industrial quality analysis applications data samples have small size, are coarsely discretized, and potentially contain outliers. In order to ensure optimal decisions based on that data, a robust method for parameter estimation and hypothesis testing is necessary. This paper presents a modified discretized Kolmogorov-Smirnov statistic, which allows for theoretically optimal and effective treatment of such data. The proposed method is compared to existing methods and its superiority is demonstrated by a sample from an industrial process and a Monte Carlo simulation.