If you would like to understand the impact of multiple factors on some response variable, you might need to run a large number of experiments. A saturated or screening design will allow you to quickly eliminate any factors that, by themselves, do not have a significant effect. Let’s explore this a little further.
Overview: What is a saturated design?
A saturated design is used as a screening experiment to only examine the effects of an individual factor. Information regarding interactions is lost. We also refer to these experiments in terms of resolution or confounding.
A saturated design is referred to as a Resolution III experiment. This means the main effect is confounded with the 2-way interaction, making it difficult to determine whether the effect on the response variable is due to the main effect of a single factor or the 2-way interaction. If 2-way or more interactions are important for you to understand, then you will need to run a higher-resolution experiment.
An industry example of a saturated design
In the Analyze phase of a DMAIC project, the company’s Black Belt identified 7 possible independent variables that could impact her response variable. She wasn’t sure which of the 7 were statistically significant.
Although she was interested in any 2-way interactions, she didn’t want to waste a lot of time and money exploring and analyzing non-significant factors. If she ran a full factorial experiment with seven factors and two levels for each, it meant doing 2^7 runs, or 128 runs.
She decided to use a Resolution III saturated design to screen out the factors that were not significant. That reduced her runs down to 8. After running her experiment she discovered only 3 of the 7 factors were significant. She could now run a full factorial for the 3 factors and get all the information about interactions.
Frequently Asked Questions (FAQ) about a saturated design
What is the resolution of a saturated design?
Resolution III
Why would I use a saturated design?
If you have numerous factors and are not sure which ones are significant, you might run a saturated or Resolution III experiment to eliminate the non-significant factors. This will save you time and money.
If 2-way interactions are important to me, should I still use a saturated design?
No. A saturated design is referred to as a Resolution III design, which means the main effect and the 2-way interactions are confounded. You will need to run a Resolution V or higher experiment to understand the non-confounded effect of your 2-way interactions.
