How to calculate OR for ordinal outcome that fail to fit proportional odds assumption?

[Diels_O]

Dear professors, I have a question about how to calculate OR for the ordinal outcome.

As I know, the PROC LOGISTIC can generate common Odds Ratio for the ordinal outcome, and this test should be tested using the proportional odds assumption.

But how should I manage data that failed to fit proportional odds assumption?

Can PROC FREQ generate stratified Agresti/WMW GenOR for ordinal data?

How can I get it using SAS?

Thanks to everybody! I sincerely hope to hearing from you!

[Ksharp]

Maybe @StatDave knew it .

[StatDave]

See this note. As shown there, you can use the UNEQUALSLOPES option in PROC LOGISTIC to fit a fully or partially non-proportional odds model. The procedure will automatically provide odds ratio estimates for the model predictors that are not involved in interactions, or you can use the ODDSRATIO statement to obtain odds ratio estimates for predictors whether interacting or not. I'm not familiar with the generalized odds ratio you mentioned.