Leveraged Study Design for Identifying Dominant Causes of Variation

Abstract

Extended Abstract Excessive variation in critical to quality process outputs is a common challenge in manufacturing industries. For variation reduction, most process quality improvement (variation reduction) frameworks follow Juran’s diagnostic and remedial journeys [1], that is, first using some methods to find the cause(s) of output variation (the diagnosis) and then, seeking a solution for eliminating the effect of the identified cause(s) (the remedy). Among all causes of variation, usually only a few have a big impact on the overall variability [2]. Shainin refers to them as the dominant cause(s) [3, 4]. Finding the dominant cause(s) requires a systematic strategy. The Shainin System TM [3, 5] is a coherent statistical stepwise variation reduction strategy with several problem-solving techniques. One of the techniques associated with the Shainin System TM that aims to help identifying the suspect dominant causes is group comparison , which exploits the concept of leveraging by comparing the extreme parts [5]. To do so, we select two groups of six or more (typically eight) parts, one group consisting of parts with large and the other with low quality characteristic values. Then, only for these selected parts, we measure as many suspect dominant cause input characteristic ’s as possible. If is a dominant cause, its values must be substantially different between the two groups. Shainin [3] suggests using the Tukey end-count test [6] to compare the values in the two groups. Although the investigation plan based on leveraging is an efficient way of gathering information in searching for a dominant cause using relatively small sample size, the Shainin analysis procedure is less than ideal.