topic Re: Custom DOE - factor levels in Discussions

Custom DOE - factor levels

jjhartwi — Wed, 18 Jul 2018 18:36:27 GMT

I want to create a custom DOE. I have 1 discrete numeric vatiable with 3 levels (1,2,3). I have 2 covariate factors. I also has 1 continuous variable. The problem is I want the DOE with 5 levels for the continuous factor and JMP is only allowing me 2 levels. Am I doing something wrong when I specify the continuous variable or am I limited to just 2 levels?

Re: Custom DOE - factor levels

Mark_Bailey — Wed, 18 Jul 2018 18:55:35 GMT

Use the discrete numeric factor to define a continuous factor with pre-determined levels. It will be modeled with continuous fixed effects. The algorithm requires up to a fourth-order term to work but the estimability will be set to 'if possible.' Just leave all that alone.

Re: Custom DOE - factor levels

Dan_Obermiller — Wed, 18 Jul 2018 19:27:07 GMT

To expand on Mark's answer, a continuous variable will have the number of levels chosen to estimate your given model. If you only specify a linear term for your continuous variable, then you only need two levels to estimate that term. Any more terms would not be as efficient and less than optimal. You use Discrete Numeric whenever you want to force certain levels of a continuous factor.