nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Post-hoc test after contingency table

Jolien

Is it possible to do a post-hoc test in JMP when Chi2 of your contigency table indicated that their is a difference between two categorical data?

I have three treatments and 5 categories and I want to know if for instance there are significant differences in my 3 treatments between the categories (like categorie 1 have significant more samples from treatment A compared to treament

Is this possible in JMP?

Thx

Accepted answers

Mark_Bailey

I think that this situation will be addressed in JMP Pro 14. That is, you will be able to fit ordinal logistic models in Generalized Regression platform and use custom post-hoc tests.

All comments

Mark_Bailey

You might be able to use the custom test or contrast options in a generalized linear model fit. See Help > JMP Help and search "custom test." Select "Custom Test Option" to see the details for GLM.

Jolien

Thanks for your reply.

When I use the fit model option, fill in my two nominal response variables and choose for GLM it gives me the warning of bad data, the only option is nominal logistics, which don't give me the tests I want.

Is there maybe another option to test two nominal response variables?

Mark_Bailey

It sounded as if you had one response (five nominal categories) and one factor (tthree reatments). Why do you have two response data columns?

Also, use the Generalized Linear Model as I told you instead of the Nominal Logistic fitting personality.

Jolien

Yes indeed I have one response with five categories and one factor with three treatments.

But I do not know how I have to fill this in 'fit model'? If I take my response as Y and my factor to constuct the model I can only take GLM if I say that the distribution is binomial and that is not. So I'm stuck in filling out the 'fit model' option.