Graphical tests of checking proportional odds assumption?

[Sonnen Blume]

Hi,

I just came across the idea of using graphical tests to check the proportional odds assumption for ologit models, and wondering if this is possible in Stata as well!

Here is the link: https://stackoverflow.com/questions/...-ordered-logit

Please let me know.

Thanks.

[Maarten Buis]

There are a couple of problems with the suggestion in the link: The most damning problem is that that suggestion does not test the proportional odds assumption. The second problem is that even if we fix that display it is still no good: because now we need to judge whether two curved lines are shifted horizontally, and we humans aren't made to be good at that.

I like graphical tests of model assumptions (I am a great proponent of the inter-ocular trauma test), but this one needs some more thought before it can become useful.

[Frode Andre]

Hi,

I just came across the idea of using graphical tests to check the proportional odds assumption for ologit models, and wondering if this is possible in Stata as well!

I would use the term "Graphical illustration of the proportional odds assumption".

You may find an illustration of a similar approach using Stata in chapter 5.6 “The parallel regression assumption” in the book “Regression models for categorical dependent variables using Stata” (2001) by Long and Freese. The authors refer to a do-file from the package spostst4. However, I am not sure this is available anymore, since there has been an update to spost13-package and a new edition of the book.