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
- :
- Discussions
- :
- Fractional factorial with 3-level factors

Topic Options

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

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

Mar 24, 2010 11:54 AM
(2253 views)

I have a product mix with three components I am trying to minimize while maximizing my response. I have one response and four continuous factors. 3 of the factors are components of the product and the fourth factor is the starting temperature of the test that yields the response.

I would like to perform a fractional factorial rather than the full factorial method. I want all four factors to have three levels but reduce the number of trials from 81. A Screening design only allows 2-level factors. A Taguchi design allows 3-level factors but doesn't seem to be the right modeling method for my project.

Can a 3x3x3x3 fractional factorial be done using JMP6 or JMP7 to do this?

I would like to perform a fractional factorial rather than the full factorial method. I want all four factors to have three levels but reduce the number of trials from 81. A Screening design only allows 2-level factors. A Taguchi design allows 3-level factors but doesn't seem to be the right modeling method for my project.

Can a 3x3x3x3 fractional factorial be done using JMP6 or JMP7 to do this?

2 ACCEPTED SOLUTIONS

Accepted Solutions

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

Mar 24, 2010 4:13 PM
(2283 views)

If the three components as you mention are truly components in a formulation then I would simply use the Custom Design capability in JMP to set up a 4 factor design where the first three factors are mixture components and the 4th factor is a continuous factor = Temperature. Adding the interactions to the model and specifying some center points (N=3) the number of runs required would only be 15. By treating this as a two level design with center points you get your three levels as you desired.

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

Mar 25, 2010 5:18 AM
(2282 views)

In JMP5, the answer is YES, you can get a fractional factorial of a 3**4 design (so I assume later versions of JMP can do this also).

The fractional factorial you are interested in is indeed the Taguchi 9-run design or the Taguchi 27-run design. These are identical to a standard fractional factorial. No difference. Go ahead, use it without qualms.

I also point out that a central composite design (response surface design) often has better statistical properties than the fractional factorial of a 3**4, and usually involves fewer runs as well.

Message was edited by: Paige

The fractional factorial you are interested in is indeed the Taguchi 9-run design or the Taguchi 27-run design. These are identical to a standard fractional factorial. No difference. Go ahead, use it without qualms.

I also point out that a central composite design (response surface design) often has better statistical properties than the fractional factorial of a 3**4, and usually involves fewer runs as well.

Message was edited by: Paige

3 REPLIES

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

Mar 24, 2010 4:13 PM
(2284 views)

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

Mar 25, 2010 5:18 AM
(2283 views)

The fractional factorial you are interested in is indeed the Taguchi 9-run design or the Taguchi 27-run design. These are identical to a standard fractional factorial. No difference. Go ahead, use it without qualms.

I also point out that a central composite design (response surface design) often has better statistical properties than the fractional factorial of a 3**4, and usually involves fewer runs as well.

Message was edited by: Paige

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

Mar 25, 2010 5:33 AM
(2091 views)

Thank you Lou and Paige. I will try them out.