regression
Multinomial logistic regression
Model an unordered categorical outcome with three or more levels.
Requirements
- 3+ level outcome
- one or more predictors
- adequate observations per category
Avoid when
- the outcome is ordinal
- categories are sparse
- observations are clustered
Formula
log[P(Y=k)/P(Y=ref)] = Xβₖ
Interpretation
Relative risk ratios compare each outcome category with the displayed reference category.
Worked example
Model membership in no response, partial response, or full response from patient predictors.
Open example datasets