Is it possible to choose which fraction in a fractional factorial design?

[#1]

Hi,

I was using Easy DoE to prepare for a simple experiment and wanted to use the current process as one of the trials, a baseline.....

My inputs are as follows:

new inputs.PNG

The design, based on my requirements to have a minimum number of trials is seen below.

Inputs.PNG

Design.PNG

Unfortunately, this design does not include my current process (1000 1400 30).

Is there a way to use the other fraction of this design, which would have included my baseline process? 

- Philip

[#2 · Accepted solution]

Hi @jpol,

Which platform did you use to generate your design ? Custom Design ? I'm not familiar with "Easy DoE", I think the "Custom Design" runs in the background in most cases but not sure.

Using the platform "Screening Design" (available in DoE, Classical, Two-level Screening, Screening Design) and specifying the factors and ranges as you did, I'm able to have a design respecting the inclusion of your "forced run" by selecting the option "Construct a main effect screening design" with minimum number of runs (4) :

Victor_G_0-1697638558892.png

Proposed design (1st run does correspond to your current process conditions, DoE attached):

Victor_G_1-1697638642566.png

If you want to be sure the platform will always generate this design (as the other "half" can also be proposed), you can click on the other option "Choose from a list of fractional factorial design" and select the first one with run size of 4 :

Victor_G_2-1697639126854.png

After clicking on "Continue", go into the panel "Display and Modify Design" and in the "Change Generating Rules", uncheck "+/-" and apply.
This will force the design creation to have a row with only minimum values (a row with "---"), corresponding to your current process settings (you can check the coded levels design in the related panel) :

Capture d'écran 2023-10-18 162607.png

Hope this will help you,

[#3]

Thank you Victor for your prompt and detailed response.

This does exactly what I wanted to do. 

BTW  Easy DoE is a new platform in JMP 17.See details below.

Easy DOE (jmp.com)

- Philip

[#4]

Hi @jpol,

You're welcome, glad this solution was helpful for you.

As an additional information, when you left the "+/-" checked you obtain an Hadamard matrix, when unchecked, you have the equivalent Hadamard matrix (but negating all signs in the matrix compared to the previous one found).

If you check here (https://en.wikipedia.org/wiki/Hadamard_matrix) you can see that the design proposed (with unchecked "+/-") corresponds to the opposite of matrix H4 (randomized), where you take the opposite of all signs (except the first column used for calculation of the intercept, so it stays at +1).

I know Easy DoE, but I'm not using it (apart for teaching/demonstration because some of the graphs are very interesting, like the Response Strength plot), since I prefer being able to choose (several) platform(s) depending on my needs and constraints :)

Best,

Victor