Ordinal Logistic Regression

[Eugenia Lee]

Hi all. I have questions about ordinal logistic regression model when weights are used. The proportional odds assumption was not met, so I used [gsvy:gologit2 with autofit option] to fit my model. How do I check for goodness of fit in this case? Most commands don't seem to work with weights. Also, how can I check for linearity assumption in this case? Thank you so much for your insights.

[Carlo Lazzaro]

Eugenia:
welcome to this forum.
Have you already read Richard Williams 's teaching notes (gologit2 manual) on his community-contributed module (as the FAQ kindly request you to mention it) -gologit2-?

[Eugenia Lee]

Hi Carlo, thank you for your post. I have read the teaching notes, but I don't see a discussion on how to check for the goodness of fit and linearity assumption when running ordinal logistic regression with weights.

Best,
Eugenia

[Richard Williams]

Svy commands in general can be a little daunting because so many things we are used to doing -- including many diagnostic and GOF tests -- do not work with svy:. See my handout

https://www3.nd.edu/~rwilliam/xsoc73...yCautionsX.pdf

From within Stata I suggest you check out

help svy postestimation

and see if anything there will help you.

Also helpful may be

https://www3.nd.edu/~rwilliam/xsoc73994/Margins05.pdf

which shows aids to interpretation.