Berg, Douglas (1991, ASQC) Hydra-matic Division, General Motors Corp., Ypsilanti, MI
This paper reviews the theory and concepts of "components of variance." The author stresses the importance of understanding the model behind an experiment, analyzing the relationships among the factors in an experiment, using graphics to test the data against the model, and examining Taguchi methods to understand the sources of variation.
The experimental model is subjective; however, it is essential for its ability to relate observations to the effects of various factors and their interactions. Selecting the number and nature of the factors to be included in the experiment requires subject matter expertise and judgment. The product or process should be analyzed based on the data and then projected, with possible changes, into the future. This analysis requires more than the analysis of variance (ANOVA) tables. Plots and graphs are also essential for checking assumptions, analyzing data, and presenting the results to decision makers.
Reductions in variation correspond to some measure of quality or customer satisfaction, which can be balanced against the cost of implementing improvements. When practitioners have pertinent information concerning the state of the system during the experiment and its current state, their decisions are more reliable due to greater assurance or confidence in the results of the experiment.
Automobile industry,Manufacturing,Statistics,Taguchi method,Analysis of variance (ANOVA)