If you need to experiment and are dealing with a hard-to-change factor, consider using a split-plot design. Split-Plot experiments can significantly speed up data collection that would otherwise take a prohibitively long time, or even be impossible.
To Demo a Split-PlotDesign in JMP Let's consider the Manufacturing challenge of the effects of five factors on the thickness of automotive vinyl that is used to make seat covers.
In JMP we accomplish this by going to DOE menu then Custom Design.
- Double-click Y under Response Name and type thickness.
- Keep the default Response set to Maximize.
- Load the Factors by using a prebuilt Sample Data Table using the Custom Design Red Triangle and Load Factors
The Vinyl Factors.jmp sample data table is located in the Design Experiment folder. You could also just open this Factor jmp file, and select Load Factors and JMP will figure it out.
The Custom Design Window will expand down.
[On Screen - Select all 5 Factors]
- Click Interactions then 2nd.
- Click OK to dismiss the informative message.
- Enter 7 next to Number of Whole Plots.
- For Runs let's override the default and enter 28 next to User Specified.
To reproduce these exact results to what Design I'm creating here. I'm navigating to Custom Design then Red Triangle, then Set the Random Seed to 123686 and Number of Starts to 10.
Those numbers are below this video.
In constructing a design on your own, these steps are not necessary.
Click Make Design to which will reveal and expand the Design Outline.
Note that the whole plot factors, extrusion rate, and temperature, are reset seven times by the levels of the factor Whole Plots. Within each level of Whole Plots, the settings for the mixture ingredients, m1, m2, and m3, are assigned at random.
Now it time to Run the Experiment and collect our Data, and there's a premade sample data file ready linked below this video. But 1st I'll select Make Data Table.
And you can see here that the last Column Thickness, our response is blank.
I'll copy the recorded run values of thickness to that Column over linked data table. This is essentially the Collection step. Now from the Table panel, click the green triangle next to the Model script to launch the pre-built Fit Model Window
We will review the model and select Run to see the Split-Plot Analysis DOE Report
The Parameter Estimates report shows that the three mixture ingredients, as well as the extrusion rate times m3 interaction, are significant.
Under the Red Triangle, I'll select Factor Profiling then Profiler.
After Maximizing Desirability I can view an optimal factor combination at the bottom of the Profiler producing a vinly thickness around 12.2.
For more guidance on Split-Plot Designs, please follow links below this video. Thank you for using JMP.
Set the Random Seed to 123686
Set Starts to 10
REML Link
Split-Plot https://www.jmp.com/support/help/13-2/Split-Plot_Experiment.shtml
