Choose Language Hide Translation Bar
Level I

DOE - What is modeling type of the symbolized factor?


When I design DOE, there is a question about the modeling type of the factor.
The actual value of the factor to be used in th experiment is countinuous data,  but it is dymbolized so that the user can select only some specific values.
Symbolized value of the factor(x) is 0 to 16.(integer) But the actual values are decreased as an exponential function.(1/e^x)

If I use the symbolized value for factor, what type of modeling should be selected? continuous or discrete or categorical?


Re: DOE - What is modeling type of the symbolized factor?

I don't know the answer but here are some considerations to help you find the answer.


How will you model the effect of this factor? Will it produce a linear or non-linear response? Will it interact with other factors?


How many levels do you need? Do you want optimal levels for the specified model or do you want to specify the levels yourself?


I would use the Continuous factor type generally and let JMP determine the optimum levels to estimate the model parameters. I would use the Discrete Numeric factor type if I need to specify the factor levels and model a continuous factor. The Categorial factor type is also possible but based on your brief description it does not seem to be a good choice.


I do not understand your meaning of the term "symbolized" so I might be wrong about what you want to do. So the factor level is just the power of the reciprocal exponential function? I suggest that you use a p-function for the sake of the linear regression model if factor is going to change exponentially.

Learn it once, use it forever!
Level VI

Re: DOE - What is modeling type of the symbolized factor?

I'll add a question to those contributed by @markbailey :

1. From a communication and interpretation of results point of view, will the transformed values create confusion or muddy the waters wrt to the practical use of the model for decision making or explanatory insight? Many if not all of JMP's visualizations will use the values contained in the raw data table that were used for modeling as x and y axis values and if these values aren't meaningful or mask your message, then you may want to think about how you create these visualizations and share them with others. I can think of workarounds but these are contextually dependent and hard to generalize.


I agree with @markbailey from what you've shared so far I think continuing to treat the transformed values as continuous should be where you start your analysis.

Article Labels

    There are no labels assigned to this post.