Ordered Logistic Regression Tutorial
ParadaCarleton opened this issue · 2 comments
Leaving this here for someone to do in the future if they want/can -- we should probably have a tutorial for ordered logistic regressions. They're probably the most criminally underused statistical model in the social sciences. There's also a decent chance that someone will take a look at the multinomial regression and incorrectly think that's what they want to use when they have an ordered category, or worse still, a linear regression.
See also the related #67
Maybe, this could just be a note in the linear regression? Do you, also, have a citation for the claim that these models are so great? Out of curiosity.
Maybe, this could just be a note in the linear regression? Do you, also, have a citation for the claim that these models are so great? Out of curiosity.
Ordered logistic regressions are very different from linear regressions; I don't think they should be lumped together. It's less that these models are great, and more that they're a decent model for analyzing ordinal data. Psychologists use linear regressions to model ordinal data (incorrectly) all the time instead, which creates some pretty severe noise and puts a downward bias on their regression coefficients. See here.