Be Sure to Check the Residuals When Analyzing the Outcome of your Designed Experiments

Be Sure to Check the Residuals When Analyzing the Outcome of your Designed Experiments

Continuing his series of tips for people implementing Continuous Process Improvement, our tutor Albert Plant writes:

There is a natural tendency for experimenters to hurry to look at the effects graphs, ANOVA’s and other analysis, when they have completed their hard work running experiments, and they may forget to check the residuals. Residuals are unexciting, but contain crucial evidence on the validity of the experimental outcome. I have seen experimenters writing up lengthy reports on the outcome of their experiments, only to discover later that the condition of the residuals indicates that they have drawn incorrect conclusions. For example, an analysis of the residuals may indicate that it is necessary to transform the data to obtain the most appropriate model.

All DOE computer software, such as Minitab, provides extensive analysis of DOE residuals. Design Expert DOE computer software has no less than eleven separate graphs plus additional text output analysing the residuals. Clearly, the people who develop the mathematical tools for us to design and analyse our experiments are in no doubt as to the importance of the residuals in DOE.

Learn more about how design and analyse experiments and how to get the most information from the residuals, by attending our Design of Experiments training courses.