Analyzing Unassigned Interactions to Strengthen DOE Strategy 2004-01-1746

Low resolution fractional factorial experimental designs, used in screening, are more popular than ever due to the ever increasing costs of materials and machine time. Experimenters have to be more precise in their analysis, making every degree of freedom count. Resolution III designs are becoming more commonplace for use in screening designs. When running unsaturated resolution III designs there are extra degrees of freedom stemming from unassigned interactions. It is common practice to utilize these extra degrees of freedom to approximate error. In many cases, this common practice can over state the error and lead to erroneous results regarding factor statistical significance. Utilizing saturated resolution III designs and statistically analyzing unassigned interactions while estimating the error with replication is a method for strengthening the DOE strategy and improving the results from screening designs.