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
- :
- Mixed-level Custom DOE

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

Jul 2, 2010 10:01 AM
(1019 views)

I'm new to JMP and have very little experience with statistics.

I'm a graduate student and I'm creating an experiment with 3 5-level factors and 3 2-level factors. The custom design in JMP yields 103 systems out of the possible 1000 systems. Due to the large amount of simulation for these systems, 103 is an acceptable number. I am trying to capture all the main effects and 2-way interactions.

I assume this is some sort of mixed-level fractional factorial design but I am curious as to how a number like 103 would be generated. From my understanding of fractional factorials (ie: 5^(3-1)*2^(3-1) = 100 systems), a number like 103 could not be generated.

Any help would be greatly appreciated.

I'm a graduate student and I'm creating an experiment with 3 5-level factors and 3 2-level factors. The custom design in JMP yields 103 systems out of the possible 1000 systems. Due to the large amount of simulation for these systems, 103 is an acceptable number. I am trying to capture all the main effects and 2-way interactions.

I assume this is some sort of mixed-level fractional factorial design but I am curious as to how a number like 103 would be generated. From my understanding of fractional factorials (ie: 5^(3-1)*2^(3-1) = 100 systems), a number like 103 could not be generated.

Any help would be greatly appreciated.

1 REPLY

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

The minimum number of runs will generate a saturated design. That is to say, the number of runs corresponds to the number of terms in the regression model required to estimate the main effects and interactions.

You can proceed to make the design and add some simulated responses (either use the formula editor or use the simulate responses option on the custom design menu) then run the model script on the design table - you will see that the model is saturated (no degrees of freedom to estimate statistical significance) and you will see a list of all the terms in the model (you can count them to convince yourself of the minimum number of runs required).

In general computer generated designs (e.g. D-optimal) don't really care about symmetry and balance in the outcome - they just look to optimise the design matrix (subject to a definition of "optimise"). JMP applies some heuristics to ensure that the default recommendation has runs in addition to the minimum number and exhibits a degree of "balance".

You can proceed to make the design and add some simulated responses (either use the formula editor or use the simulate responses option on the custom design menu) then run the model script on the design table - you will see that the model is saturated (no degrees of freedom to estimate statistical significance) and you will see a list of all the terms in the model (you can count them to convince yourself of the minimum number of runs required).

In general computer generated designs (e.g. D-optimal) don't really care about symmetry and balance in the outcome - they just look to optimise the design matrix (subject to a definition of "optimise"). JMP applies some heuristics to ensure that the default recommendation has runs in addition to the minimum number and exhibits a degree of "balance".