2
$\begingroup$

Quick question, is ordinal logistic regression a linear or nonlinear model? Finding different sources supporting the other, and the more I read the more I get confused myself.

Perse, it should fall under GLM, as logit does. Or?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
5
$\begingroup$

All ordinal regression models in use are nonlinear because they are anchored to a probability scale and regression effects must bend to keep probabilities within [0,1]. You could say that the cumulative probability model class of ordinal regression (proportional odds, proportional hazards, probit) falls under GLM, though most people's concept of GLM includes only one intercept and cumulative probability models have $k-1$ intercepts for $k$ distinct values of $Y$. The software needed to fit ordinal models is specialized and is not handled by GLM software, because of the multiple intercepts. Some resources are at https://fharrell.com/post/rpo.

Improve this answer
$\endgroup$
3
  • $\begingroup$ Thank you @Frank Harell. $\endgroup$
    – Moa Herrgård
    Commented Apr 26, 2022 at 14:41
  • 2
    $\begingroup$ On this site, thanks is expressed by upvoting an answer or by selecting it as "the answer". $\endgroup$
    – Frank Harrell
    Commented Apr 26, 2022 at 15:25
  • $\begingroup$ If you found this answer helpful, then please consider upvoting and/or accepting it. $\endgroup$
    – kjetil b halvorsen ♦
    Commented Apr 26, 2022 at 16:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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