Section 8 Ordinal Outcomes

This section will cover ordinal logit and probit regression. These are models that work well for outcome variables that include a set of (more than 2) ordered discrete categories.

Often, survey responses have this form of outcome (e.g., a likert scale from “strongly agree” to “strongly disagree”). We also might categorize behavioral responses in an ordinal way from “stay home” to “protest”, for example.

You can review the following for additional external resources on this section.

  • King, Gary. 1998. Unifying political methodology: The likelihood theory of statistical inference. University of Michigan Press. Chapter 5.4.
  • Resources for ordinal models in R here and here
  • A section from Gelman and Hill Chapter 6, posted to Canvas.