Document Type


Publication Date



A company manufacturing fans wished to conduct an experimental design to determine the best combination of three factors affecting the breaking torque of the fans. Analysis of the data ceased to be straightforward when the authors found that the data failed the test for homogeneity of variances. After unsuccessfully attempting to transform the data and thereby meet the assumptions necessary to carry on the analysis, the authors relied upon a graphical analysis and a careful study of the means for each design point. This paper describes a statistically sound but novel strategy used to complete the analysis.