Section 9 Multinomial Outcomes

This section will describe cases where we have unordered categorical outcome variables.

The following supplemental resources may be useful here.

  • Rodriguez Chapter 6.1-6.2.4 (see pdf on Canvas)
  • Ledolter Chapter 11, pgs. 132-134 (see pdf on Canvas)
  • Multinomial R resources via UCLA

\(Y_i\) can take any of \(J\) categories like ordinal logit. Usually, we use this for more than 2 categories, or else it would collapse to a simple binary outcome. The key here is these are not ordered!

  • For example, a nominal set of categories might be vote choice between different candidates,
  • Instead, it could be choice of different investment or insurance plans,
  • Perhaps it is a choice of locations to target in an attack,
  • Or maybe it is a type of campaign strategy– direct mail vs. online ads, etc.