Changing Design Roll after making Custom DOE table

[#1]

Is it allowed to change Design Roll in column property from Discrete Numerical to Continuous in Custom DOE table?

If I make Custom DOE with Discrete Numerical factor, no wonder I can select only discrete value when analyzing response. However continuous value becomes available when I change Design Roll into Continuous. Prediction expression of the model does not change and response is consistent. I would like to know if there is any issue in this approach or not.

[#2 · Accepted solution]

Hi @RF,

For discrete numeric factors, you can estimate main effects, interactions, quadratic effects... same as for continuous numeric factors. So the difference in modelisation/design creation is not big between these two types. As you said, it will only constraint your analysis, by only being able to select/have discrete values for this factor.

Changing a discrete numerical factor to a continuous one should not be a problem, since you had at least 2 discrete levels, which is required for the analysis of a continuous factor (level -1 and level +1).
You might even have more information if you had previously more than 2 levels in your discrete numerical factor : it should be possible to test for quadratic effect through a lack-of-fit test (in case of middle value for this previously discrete numeric factor).

I don't know how accurate my answer will be on your use case, but I think there should be no problem on a general view. 
You may be interested in this discussion : Solved: Continuous Vs Discrete Numeric Factor - JMP User Community

[#3]

Thank you Victor for your explanation and reference discussion. I will try this approach.

[#4 · Accepted solution]

Yes, it is allowed. The choice of discrete numeric factors is how you tell JMP that you will only test a continuous factor (yes, it is continuous) at certain levels. Then JMP adds terms to the model so that the coordinate exchange algorithm will keep those levels in the design. It is very important in the design of the experiment. It is not important at all in the analysis.

[#5]

Thank you Mark for your answer, I will try.