BookmarkSubscribe
Choose Language Hide Translation Bar
DS
DS
Community Trekker

Definitive screening design blocking

Hi All,

 

  I have some colleagues that are considering running a definitive screening design for a pilot plant production scale run, and have asked for my help in creating a DOE for them.

 

  The pilot plant runs on a day/night shift fashion. What I'm curious about is if it's possible to get JMP to for the output DOE table to alternate runs in a day/night fashion. I can block or add a categorical factor in there to represent the day/night shifts, but I can't get JMP to force the output table to alternate based on that column. I can always do it manually with a script or via a column formula, but wondering if there is already some built-in method to do this.

 

Thanks!,

DS

0 Kudos
2 REPLIES 2

Re: Definitive screening design blocking

What version of JMP are you using?

 

You could create an indicator column separate from the design and use a sort to achieve the order that you want.

 

I used JMP 14 and the blocking option in DSD platform. I created a DSD for 5 continuous factors. (I know, not really a screening situation.) Here is the result without blocking:

 

Screen Shot 2019-03-18 at 5.20.26 AM.png

 

Then I decided to create 4 blocks for 2 shifts over 2 days. I want to be able to estimate the potential quadratic terms. Here is the result:

 

Screen Shot 2019-03-18 at 5.21.21 AM.png

 

Here it is the same design with the Randomize within Blocks option selected:

 

Screen Shot 2019-03-18 at 5.23.24 AM.png

 

Is this example not what you are trying to do?

Learn it once, use it forever!
0 Kudos
DS
DS
Community Trekker

Re: Definitive screening design blocking

Hi @markbailey,

 

  I'm using JMP Pro 14.

 

  Thanks for the reply. The output you showed isn't quite what I'm after. I think I was able to get it to work, although perhaps in not the most elegant way. I created a new column formula: Modulo(Row(), 13)) and then sorted the table ascending by that new column. In this case, the 13 comes from the number of runs/2 (2 shifts). I get 26 runs simply because I put in 10 factors and added two blocks (day shift and night shift) as well as center points for estimating curvature.

 

Here's the original table randomized within blocks and the last column is my formula from above.

orig.png

 

I then sort this table according to that column's values and get the output below. Here, my runs alternate by block: day shift = block 1, night shift = block 2. That way, each shift can modify their settings accordingly. This is ultimately what I want. I'm sure there are more streamlined and elegant ways (e.g. from a JSL viewpoint) to do it, but this works, albeit more of a brute force method.

mod.png

 

If I would do four blocks as in your example, then I would change the formula to Modulo(Row(), 7) so that it created rows numbered by the blocks: 1, 2, 3, 4, and then repeat that seven times.

mb.png

  

This is conceptually what I'm after. And, although this method works, it's more of a force it to fit approach. Could be nice if this can somehow be an option within the DOE table output choices.

0 Kudos