Ordinal regression Part 3: Proportional odds assumption
HTML-код
- Опубликовано: 22 авг 2024
- The last video of the series discusses what we mean by the proportional odds assumption when we conduct ordinal regression models. It illustrates the assumption by showing an example comparing results from a multinomial and an ordinal model.
This video (presented by Dr Heini Väisänen) is part of an NCRM Online Resource (including slides, worksheet, data and recommended literature). To see the full resource please visit www.ncrm.ac.uk...
Please note: we may be unable to respond to individual questions on this video.
The National Centre for Research Methods (NCRM) delivers research methods training through short courses and free online resources.
- Visit the NCRM website: www.ncrm.ac.uk
- Browse our short courses: www.ncrm.ac.uk...
- Find online resources: www.ncrm.ac.uk...
Follow NCRM on social media:
- Twitter: / ncrmuk
- LinkedIn: / ncrmuk
- Facebook: / ncrmuk
- RUclips: / ncrmuk
Subscribe to the NCRM monthly newsletter: www.ncrm.ac.uk...
Thank you so much! The series of videos do help me a lot!
Wonderful explanation, thank you so much!
Hi, the probability of falling into the "very likely" category for the students with educated parents is actually .2100931 using the Ordered logistic regression instead of 0.25.
Thank you for your help!!
Thank you very much. Very clear and helpful!
Thanl you, clearly explained
A solution for a ordered logistic regression model that don’t respect the assumption of proportional odds would be using a Unconstrained Generalized Ordered Logit Model?
How can we do ordinal regression where the dependent variable and independent variables have large number of items whose observation is collected in likert scale? can we use the transformed mean for analysis of ordinal regression?
Do you have an example with interval predictive variables? Such as predicting future grades based on standardized assessments.
how we can calculate the cumulative odds ratio for Ologit
If the assumption holds, and if the OR between the highest (ref value) and the 3 other categories is 2.5, then does that mean the OR is also the same for category 1 + category 2 v. 3 and 4? And category 1, 2, 3 v 4. They are all approximately 2.5?
what if some of our independent variables met PO assumption b and some not
What if only one (out of 9) of our variables doesn't pass proportional odds assumption (brant test in R, p-value in this test 0.04)? Should we resign from conducting ordinal regression and choose MLR instead or exclude this variable although it's significant in the model (p - value 0.0003) or are there any other ways do deal with that?
Hi, you talked about the outcome variables. If outcome variables are greater than 2, you can use Ordinal Model. What if there is only 1 or 2 outcome variables? Can you elaborate this further. Thanks
Try using a binary logit model instead of an ordinal model
Can someone help please? How do I do interaction (like in anova) in ordinal regression? No information on this anywhere :(