I am desperately looking for a way to build a multilevel ordered regression for variables which showed to violate the proportional odds (parallel lines) assumption in a single level model variant.

For single level model building I used the gologit2 package in Stata, which relaxes the assumptions. The author says there is no multilevel support for this package.

I found a very old package called gllamm that looked promising. However, a) it is not suported anymore that leads to b) factor variables cannot be used.

Does anyone know a package preferrably in Stata that can be used to solve my problem?

If not Stata, maybe something in R?

Thanks a lot!


You can use gllamm in Stata for this. It is a user-written program that is still widely used, so I wouldn't hesitate to employ it for these purposes. You can find information on fitting such a model here (zip file with presentation, syntax, and data from a talk Sophia Rabe-Hesketh gave at a 2009 Stata conference). Simply install gllamm using ssc install gllamm from Stata. I ran the proportional and non-proportional odds models in the do file at that link and it works as advertised in Stata 15 MP. In terms of factor variables, you can just create dummy variables yourself for the variable of interest. E.g., tab factor, gen(fac_lev) will generate a set 0/1 dummies you can use in your model.

In terms of R, you can use mixor(), which was developed by Archer, Hedeker, and others to fit single or multilevel and longitudinal ordinal response models. See the package vignette here.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.