topic Re: Ordinal logistic regression in Discussions

Ordinal logistic regression

Ina123 — Fri, 09 Jun 2023 00:37:09 GMT

I am using JMP (Fit Model) to do a ordinal logistic regression.

Now I am looking for a way to do a test the assumption that "proportional odds" are involved in my model.

Does anyone know how to do this? Or is there a possibiltity that JMP does this kind of test (also referred to as "Approximate likelihood-ratio test of proportionality of odds" or "Brant Test of Parallel Regression Assumption" or "score Test for the proportional odds assumption" automatically before running the model?

Any help would be much appreciated!

Re: Ordinal logistic regression

Mark_Bailey — Thu, 05 Aug 2021 14:20:31 GMT

You might use the difference between the 'cumulative logits' behind the ordinal logistic regression model and the 'generalized logits' behind the nominal logistic regression model. The ordinal logistic model is more restrictive, so the same coefficients apply. The only difference between levels is the intercept value. On the other hand, the nominal logistic model uses a full set of parameters for each level.

You might compare the results of the ordinal logistic regression to the results of the nominal logistic regression to determine if the restriction is valid.