Fractional Designs

In this example, we have not considered any practical limits on the number of experimental runs to be performed. This is not the case for all experiments, particularly those with more factors or levels, or those with time constraints or with scarce or very costly materials.

To save time or materials we might run fewer replicates, but then we'd lose the advantages of replication. Instead, if we can confidently forego the testing of all possible interactions we can reduce the total number of runs by using a fractional design. In a fractional design we deliberately omit some interaction combinations in exchange for the savings in experimental runs. We plainly collect less information than with a full factorial design, but we can still obtain useful data that we can analyze fruitfully.