Discriminate analysis with nominal variables

tuo88138 — Fri, 09 Jun 2023 00:58:40 GMT

Hi, I have nominal variables and I want to do a discriminate analysis. when I run the algorithm it gave me the below error. However, I pressed the continue and I got an answer. Since I got the below error, Can it be considered a correct answer?

Re: Discriminate analysis with nominal variables

Ross_Metusalem — Wed, 14 Dec 2022 20:33:14 GMT

The short answer: I don't think you want to trust those results. Only continuous variables should be used as Y variables in discriminant analysis.

The longer answer: Discriminant analysis requires continuous Ys and a categorical X. In JMP, data table columns have both a data type and a modeling type (see here for the difference), and it looks like you have some columns with a numeric data type but a nominal or ordinal modeling type, meaning that while they're represented by numbers in the data table, they're really not continuous variables. Sex could be a good example: It may be represented as 0 and 1 in the table but really is nominal (male = 0 and female = 1). The discriminant analysis launch dialog will allow you to input Sex as a Y variable due to the numeric data type but thankfully throws up this warning if Sex is set to a non-continuous modeling type. (If you were to change Sex's data type to character, then the launch dialog would reject it outright when you try to put it in the Y role.) Including a non-continuous numeric column as a Y in discriminant analysis should be done only when you believe the variable can be treated as a continuous variable for the purposes of this specific analysis.

Re: Discriminate analysis with nominal variables

tuo88138 — Tue, 20 Dec 2022 17:29:53 GMT

Thank you so much

topic Discriminate analysis with nominal variables in Discussions

Discriminate analysis with nominal variables

Re: Discriminate analysis with nominal variables

Re: Discriminate analysis with nominal variables