Choose Language Hide Translation Bar
Highlighted
Level I

## DOE Design with multiple factors

I have a question on how to design a DOE:

1. Multiple groups of factors; each group with multiple factors;

2. There are common factors be shared cross groups;

3. How to desing DOE to minimize runs--no full factoria design;

Example:

A1/A2/A3/A4/A5/A6 as group # 1;

A1/A2/B8/B9/B10 as group # 2;

A1/ A3/ C15/ C16/ C17 as group # 3;

Is there a way to design DOE together like:

A/A2/A3/A4/A5/A6 + B8/B9/B10--please do not put all factors into screening design;

Is there any Pto and Con that you foresee?

4 REPLIES 4
Highlighted
Staff

## Re: DOE Design with multiple factors

We will deal with the number of runs later. Please explain the nature of a 'group.' What constitutes a group? Why are the factors grouped?

I do not understand your example. Are A, B, and C factors? Is a symbol like A1 mean factor A at level 1?

Learn it once, use it forever!
Highlighted
Level I

## Re: DOE Design with multiple factors

Mark,

There are multiple compoments in one complex assmelby which some components are common which are related with other components.

A1/A2... are factors, each with multiple levels;

B8/B9... are factors too, each with multiple levels s well;

Each group of factors will be evalauted with an individual resposne.

Highlighted
Staff

## Re: DOE Design with multiple factors

So does each group represent the components within a sub-assembly? What are the factors? Different types of component?
Highlighted
Level VI

## Re: DOE Design with multiple factors

Would you be able to create a table of all the possible experimental combinations, that exclude the combinations that you cannot do or that you are not interested in?

If so, you could then take that "full" design and find an optimal subdesign of a given number of runs in Custom Design platform.  You would need to create the table, then run Custom Design with that table as the current data table, and add all the factors as covariate factors.  You then choose the kind of model you want to fit to the data after the experiment is done, and the number of runs, and JMP will search out the optimal design from list of candidate points.  There is also a way to ensure that certain combinations are included in the design by preselecting the rows in the full table and check the option to include those in the Custom Design platform.

I recently posted how to do this to find an optimal balanced incomplete block design (https://community.jmp.com/t5/Discussions/Balanced-Block-Designs-Using-Covariate-Factors-in-Custom-De...), the approach described above is the same.

Article Labels

There are no labels assigned to this post.