Page 458 - Design for Six Sigma a Roadmap for Product Development

Fundamentals of Experimental Design 417

may indicate that we may have missed important factors in the experi-
ment. DOE data analysis can identify significant and insignificant
factors by using analysis of variance.
2. Ranking of relative importance of factor effects and interactions.
Analysis of variance (ANOVA) can identify the relative importance of
each factor by giving a numerical score.
3. Empirical mathematical model of response versus experimental
factors. DOE data analysis is able to provide an empirical mathe-
matical model relating the output y to experimental factors. The
form of the mathematical model could be linear or polynomial, plus
interactions. DOE data analysis can also provide graphical presen-
tations of the mathematical relationship between experimental fac-
tors and output, in the form of main-effects charts and interaction
charts.
4. Identification of best factor level settings and optimal output per-
formance level. If there is an ideal goal for the output, for example, if y
is the yield in an agricultural experiment, then the ideal goal for y
would be “the larger, the better.” By using the mathematical model
provided in paragraph 3, DOE data analysis is able to identify the best
setting of experimental factors which will achieve the best possible
result for the output.

Step 7: Conclusions and recommendations
Once the data analysis is completed, the experimenter can draw prac-
tical conclusions about the project. If the data analysis provides
enough information, we might be able to recommend some changes to
the process to improve its performance. Sometimes, the data analysis
cannot provide enough information, in which case we may have to do
more experiments.
When the analysis of the experiment is complete, we must verify
whether the conclusions are good. These are called confirmation runs.
The interpretation and conclusions from an experiment may
include a “best” setting to use to meet the goals of the experiment.
Even if this “best” setting were included in the design, you should
run it again as part of the confirmation runs to make sure that noth-
ing has changed and that the response values are close to their pre-
dicted values.
In an industrial setting, it is very desirable to have a stable process.
Therefore, one should run more than one test at the “best” settings. A
minimum of three runs should be conducted. If the time between actu-
ally running the experiments and conducting the confirmation runs is
more than a few hours, the experimenter must be careful to ensure
that nothing else has changed since the original data collection.