- Multinomial logit is a set of binary logit models for all pairs of outcomes
- Multinomial logit can be overwhelming if you try to interpret in terms of the coefficients for outcome contrasts
- The same tools for interpretation in binary logit can also be applied to multinomial logit
- Predicted probabilities and MEs greatly simplify interpretation
- These are usually what you care about anyway
- Plots of effects are often necessary to understand the patterns
- Always make the plot to guide your understanding regardless of whether it ends up in the published paper
- These also help reviewers understand the model results
Back to top