Turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

- JMP User Community
- :
- Discussions
- :
- Is there a better DOE design for me?

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

Highlighted

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Sep 21, 2011 12:47 PM
(6482 views)

Hello everyone,

I'm a novice JMP user and thanks in advance to any advice you can give!

I'm interested in the main effects and two way interactions for four, continuous, 2 level factors in my process. Using the Custom Design to create my DOE, I get a minimum of 11 runs. Each run will likely take a couple of hours for the process to settle out after a change. If I assume my current process operates at the center points (this may or may not be a good assumption), I can add 3 center points runs with minimal impact to the overall experiment time. A center point run going into and out of the experiment, and one somewhere during the experiment. Does this sound okay?

Is there a more efficient design to understand the main effects and two way interactions?

Thanks.

1 ACCEPTED SOLUTION

Accepted Solutions

Highlighted
Seems like a good approach. You could also consider a classical screening fractional factorial design under the screening pulldown. For 4 factors you could obtain a Resolution IV design (2-way interactions confounded with other 2-way interactions and main effects confounded with 3-way interactions.). This could be accomplished in 8 experiments and if you include your 3 center points then 11 total experiments would be required. This would save you 3 runs for verification of the model. Both approaches would work. You don't mention how much material you have and how many experiments you can afford to run. This latter point usually drives the decision. I just wanted to point out that you should always reserve some material for follow-up experiments.

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

2 REPLIES 2

Highlighted
Seems like a good approach. You could also consider a classical screening fractional factorial design under the screening pulldown. For 4 factors you could obtain a Resolution IV design (2-way interactions confounded with other 2-way interactions and main effects confounded with 3-way interactions.). This could be accomplished in 8 experiments and if you include your 3 center points then 11 total experiments would be required. This would save you 3 runs for verification of the model. Both approaches would work. You don't mention how much material you have and how many experiments you can afford to run. This latter point usually drives the decision. I just wanted to point out that you should always reserve some material for follow-up experiments.

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Highlighted
##

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Is there a better DOE design for me?

Thanks for confirming I'm on the right track, LouV. I did see the Resolution IV design while creating different designs. For my particular application, I'm confident all four parameters will impact the response. I'm just not sure of each factor's significance nor the significance of any two way interactions . I suspect interactions are present, so I'd like to evaluate those too.

I'll execute this experiment on a continuous process, so material isn't a concern. As you correctly pointed out, it's the machine time that's the deciding factor. Hopefully, by setting the factor levels properly, I'll be able to create a useful model for future improvements.

Thanks again!

Article Labels

There are no labels assigned to this post.