User Community
cancel
Turn on suggestions
Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.
Showing results for 
Show  only  | Search instead for 
Did you mean: 
  • Don’t miss your chance to experience Discovery Summit Europe at our best available rate! Early bird registration through 31 Oct.
  • Image to Table JMP extension imports single or batch images as data tables. Useful to extract curves from pictures or convert image data to curve data to use in analyses like Functional Data Explorer. Download and install the extension.

Discussions

Solve problems, and share tips and tricks with other JMP users.
Choose Language Hide Translation Bar
Vic
Vic
Level I

Discrete?Continuous? Latin Hypercude Design in JMP

Sep 20, 2019 01:37 PM (2095 views)

Hi all,

 

I have tried Latin Hypercube Design in JMP to reduce a total number of 18 experiments with 2 categorical factors each with 3 and 6 levels.  But the software does not allow me to choose any Space Filling Design methods. Is it because LHD is not originally suitable for discrete factors? If so, what methods should I go to build the reduce the number of the experiment?

0 Kudos
1 REPLY 1
Mark_Bailey
Staff Mark_Bailey
Staff

Re: Discrete?Continuous? Latin Hypercude Design in JMP

Sep 21, 2019 07:16 AM (2061 views)  |  Posted in reply to message from Vic 09-20-2019

The space filling designs are primarily intended for computer experiments (simulations) without a stochastic element. The goal is to obtain data of sufficient density to support an interpolator. Is that case your situation? These designs can be used for other purposes, though. These design methods are for continuous factors. The exception is the Fast Flexible Filling method but even it requires at least one continuous factor.

 

The minimum number of runs is determined by the number of parameters to be estimated in the model. A design with the minimum number of runs is saturated. You will have enough data to estimate all the parameters but none to test the estimates or model as a whole. A custom design requires 8 runs for the main effects model and 18 for the interaction model. I would not recommend a saturated design in most cases.

 

Can you eliminate any factor levels a priori?

Recommended Articles