Support for proportional odds ordinal regression model

[jarbet]

It would be super cool to add support for "small sample corrected cluster sandwich standard errors" for a proportional odds (PO) regression model.

The PO model works for ordinal categorical outcomes, and is also a robust alternative to linear regression for numeric outcomes in that it is robust to outliers, does not assume normality, and is more powerful for skewed outcomes. Essentially, the PO model can be thought of as an extension of the Wilcoxon rank sum test that allows adjustment for covariates:

https://www.fharrell.com/post/rpo/

This issue is similar to #65 , however, I would recommend supporting rms::orm over MASS::polyr. The author of rms::orm claims it is more computationally efficient for continuous outcomes or ordinal outcomes with many levels. I don't think MASS::polyr even supports numeric outcomes but requires the outcome to be a factor.

There are a few GEE packages for PO models, but as far as I know, none of them support small-sample cluster standard errors:

https://cran.r-project.org/web/packages/repolr/ https://cran.r-project.org/web/packages/multgee/ https://rdrr.io/cran/geepack/man/ordgee.html