Analysis of Variation Transmission in Manufacturing Processes—Part II

Abstract

Methods for analysing variation in multistage manufacturing processes in order to identify stages which contribute most to variation in the final product are a valuable prioritization tool in variation reduction studies. However, when the data are observed with significant measurement error, substantial biases which mislead the investigator can result. In addition, methods of interval estimation and diagnostic model checking are needed for proper application of these methods. In this paper, we present methods that incorporate measurement error and discuss both maximum likelihood estimation and a simpler “naive” method that is much easier to implement. We then give methods of developing confidence intervals, either in the presence or absence of measurement error. Finally, we discuss techniques for model checking.