Modeling Type

[#1]

How do you know when to use an ordinal modeling type versus a nominal type. For example, If I am using numbers to label a product, in JMP I would use that as a nominal type correct?

Then if I have a certain sequence in which these products are being shown then, whatever sequence (ex. 1, 2 or 3) would be ordinal modeling type correct? Or would they both simply be nominal type?

Thanks for your help!

[#2]

Hi @lrovelo,

Welcome in the Community !

The difference between these two types is the notion of ordering.

• In nominal type, there is no relation between different numbers, it could be independant classes.
• In ordinal type, there is a relation between the numbers or categories, they are ordered 1 < 2 < 3 ... or January coming before February, ...

You can read the JMP help dedicated to modeling types here : Understand Modeling Types

So the modeling type is not linked to the numbers/classes you write, but how they are related to each others.

Hope this answer will help you,

[#3]

Victor

That is a good start, but I think the request is for something more - at least, I would like a bit more.  What modeling platforms or methods will produce different results depending on whether a variable is nominal or ordinal?  For many purposes, the results are the same - with the only difference being whether the order is automatic or the default ordering for nominal variables (which could then be manually adjusted using Value Ordering to be equivalent to the Ordinal ordering).  But I know that some analyses are actually different if a variable is designated as ordinal.  I'd like to know where that designation actually affects the kind of analysis that is done.

[#4]

Hi @dlehman1,

Without additional information from OP, this is the answer I could think of. Your interest and familiarity with modeling may not be the same as the author.
There are more subtle changes in the different platforms linked to the choice of nominal vs. ordinal :

• In modeling, the coding may be different between categorical nominal vs. ordinal. The JMP Help section has many pages related to this difference in the modeling : Models with Nominal and Ordinal Effects, Nominal Factors, Ordinal Factors, Statistical Details for Nominal Effects Coding ... 
• In DoE having an ordinal discrete numeric factor is different that having a nominal (numerical or not) factor, as JMP can use the relation between ordinal levels to reduce the number of runs needed (you can see this as a kind of "internal" constraint)

You also have to take into account the platform used, as the coding of nominal factors can be different between "Fit Least Squares" and "Fit Generalized Regression"...

https://community.jmp.com/t5/Discussions/Random-effect-test/m-p/659523/highlight/true#M84878

https://community.jmp.com/t5/Discussions/How-DOE-analysis-handle-categorical-factor-in-regression/m-...

So it's difficult to answer generally, it's dependent on the context of the factor/variable and the goal of the study, how the analysis may make sense of the information. As you can see, it's quite easy to get lost down the rabbit hole, so I wanted to provide a short answer. If this answer is not satisfactory, then other members can jump in and OP can refine the question with more context and info to obtain more relevant and accurate answers.

Hope this answer may clarify,