Everyone Can Grow: Getting Your Hands Dirty with DOE

[#1]

At JMP, we want to shake up people’s views on how experiments should be run. Like the Tears for Fears song, Sowing the Seeds of Love, we strive to sow the seeds of love for experimentation through our software.

By learning how to do experiments more efficiently, DOE’s popularity continues to grow among industries that use JMP. Its built-in features make DOE possible (and easy), creating an ever-growing wave of new DOE enthusiasts. As @Phil_Kay and Weronika Wrazen (winner of the 2022 Let It Grow: The Garden Cress Challenge), explain in their video presentation, JMP is sowing the seeds of love for design of experiments.

Video presentation highlights

To make the concept of DOE more digestible and more easily taught to future users, a simple experiment was created to get us familiar with the concept. The purpose of the test is to maximize the growth of garden cress seeds.

Keep this in mind for later: take note of the objective of the experiment that was created. Why would someone want to measure a particular response? What could occur if an experiment’s outcomes are not fully considered? The first step in getting our hands dirty with DOE should be to define your objective.

After running the experiment for a week, Phil and his daughters were able to put their results into JMP software and do some simple analyses with the tools the software provides. They were able to conclude that the plants grown in the darker conditions were taller after seven days!

Sadly for Weronika, her first trial confirmed her own belief about not having a green thumb – the experiment failed!

She concluded that there were too many seeds in each of the boxes, the egg boxes housing the plants disintegrated, and the soil migrated into boxes that were supposed to be exclusively cotton. In her second trial, she used plastic espresso cups to keep the seeds separate.

Weronika was able to conclude that light was the most important factor in increasing height, plants that were in the dark were able to grow taller, and garden soil also produced taller plants. Although the cress that was grown in the dark was taller, it was clearly unhealthy and fragile. In contrast, the sunlight-grown cress had large, healthy green leaves. The growth of the plants in the dark seemed to stagnate in the last few days of the experiment.

Dig into these considerations about DOE

While it is important to consider the statistical numbers when running experiments, the physical outcomes that cannot be seen through data must not be ignored. Upon further exploration, it’s obvious that the roots of the plants grown in the light are much longer. If maximizing height was the main objective, rather than getting the “healthiest” plant (without even going into how you would have to quantify and define “healthy”), then you would put the seeds in the dark. When designing experiments, it is important to consider all confounding variables that could be at play in your trials and consider the risks of your objective.

Here is something I want us all to consider: would plant height be the best variable to maximize if we were trying to grow good plants? In thinking about the objective of your experiments when using JMP, you must consider your objective’s risks. Sure, the plants in this experiment were tall, but at what cost? All confounding variables must be analyzed and considered when performing trials.

To really get your hands dirty when designing experiments, think not only of what you want to maximize, minimize, or obtain. Think also of the variables (inside and outside of the statistics) that must change to reach your goal. For example, manufacturing facilities should not work at maximum capacity without consideration for their laborers. When you reach a conclusion in your experiments, dig deeper than what the numbers tell you.

At the end of Weronika's presentation, she concludes with this encouragement: never give up after your first trial. There is always something to learn from a failed experiment. By working with design of experiments, you can optimize any process and learn at a faster, more efficient pace.

Be sure to check out the original video. I hope you can leaf with some new considerations and inspirations. Happy designing!