Performance Characterization of EDA and Its Potential to Improve Decision Making in Product Batch Release

In this chapter, APSD data are examined from real products using several different strategies to compare the ability of EDA to detect APSD changes with grouped stages from full-resolution CI measurements. These comparisons were made relative to decision making associated with OIP disposition in the QC ­environment. The strategies involve (1) measurement system analysis (MSA), (2) operating characteristic curves (OCC), and (3) principal component analysis (PCA). A general description of these techniques and their basic concepts is provided in the first part of the chapter, while the computational details and results for each of the employed strategies are given following the same order in the later part of the chapter. All results point to the conclusion that compared to grouped stages, the LPM/SPM ratio from EDA is more accurate and more sensitive to APSD changes. Each of the examination strategies used different statistical methodologies, based on different assumptions, and one of the approaches used a different set of data independent of the other two. Yet the same qualitative conclusion validating the superior decision-making ability of the EDA concept was reached, regardless of approach.